# pacotes
require(tidyverse)
require(wesanderson)
require(ggridges)
require(viridis)
require(caret)
require(tictoc)
require(suncalc)
require(cluster)
require(fastDummies)
require(fpc)
require(stats)dataset <- read_csv('dataset.csv')
colnames(dataset)## [1] "X1" "Human_timestamp" "Hive"
## [4] "Apiary" "Brood_Temp" "Brood_Humidity"
## [7] "Hive_Temp" "Hive_Humidity" "Scale_Temp"
## [10] "Scale_Humidity" "Weight" "Temperature"
## [13] "DewPoint" "Pressure" "WindDirection"
## [16] "WindSpeed" "SkyCondition" "Precipitation1Hr"
## [19] "Precipitation6Hr" "Brood" "Bees"
## [22] "Queen" "Food" "Stressors"
## [25] "Space" "Code"Dessas variáveis acima, iremos somente utilizar as medições dos sensores Internos, o peso e dos sensores externos utilizaremos somente as variáveis: Temperatura, Pressão Atmosférica, DewPoint, Direção e Velocidade do vento. Incluiremos tambem as variaveis de inspeção e o timestamp para fins futuros. Não podemos incluir outras variáveis exatamente pelo fato de possuirem muitos valores faltantes. A Pressão Atmosférica Pressure poderia ser incluida porém na analise percebi que todas as medições do Apiário Beesboro estão faltando. Portanto a imputação desses valores em minha concepção geraria ruido e viés. Mas, vamos ver no que vai dar
dataset <- dataset %>%
select(Human_timestamp, Brood_Temp, Brood_Humidity, Hive_Temp, Hive_Humidity,
Weight, Temperature, DewPoint, WindDirection, WindSpeed, Brood, Bees,
Queen, Food, Stressors, Space)
dataset <- dataset %>%
filter(Bees %in% c(0, 1))
colnames(dataset)## [1] "Human_timestamp" "Brood_Temp" "Brood_Humidity"
## [4] "Hive_Temp" "Hive_Humidity" "Weight"
## [7] "Temperature" "DewPoint" "WindDirection"
## [10] "WindSpeed" "Brood" "Bees"
## [13] "Queen" "Food" "Stressors"
## [16] "Space"Vamos criar uma variável Season, para a estação do ano associada ao timestamp, isso se faz necesário já que a colméia tende a se adaptar às estações. Utilizaremos como base o ano de 2012, pois é um ano bem estruturado e que possuiu boas divisões de estações.
getSeason <- function(DATES) {
WS <- as.Date("2012-12-15", format = "%Y-%m-%d") # Winter Solstice
SE <- as.Date("2012-3-15", format = "%Y-%m-%d") # Spring Equinox
SS <- as.Date("2012-6-15", format = "%Y-%m-%d") # Summer Solstice
FE <- as.Date("2012-9-15", format = "%Y-%m-%d") # Fall Equinox
# Convert dates from any year to 2012 dates
d <- as.Date(strftime(DATES, format="2012-%m-%d"))
ifelse (d >= WS | d < SE, "Inverno",
ifelse (d >= SE & d < SS, "Primavera",
ifelse (d >= SS & d < FE, "Verão", "Outono")))
}
dataset <- dataset %>%
mutate(Season = getSeason(Human_timestamp))
colnames(dataset)## [1] "Human_timestamp" "Brood_Temp" "Brood_Humidity"
## [4] "Hive_Temp" "Hive_Humidity" "Weight"
## [7] "Temperature" "DewPoint" "WindDirection"
## [10] "WindSpeed" "Brood" "Bees"
## [13] "Queen" "Food" "Stressors"
## [16] "Space" "Season"Uma boa variável que pode nos dar grandes resultados seria o Turno do dia em que ocorreu as medições
getTurnDay <- function(data){
times <- getSunlightTimes(date = as.Date(data), lat = 35.7596,
lon = -79.0193, tz = "America/New_York")
data <- as.POSIXct(as.character(data), format = "%Y-%m-%d %H:%M:%S",tz="America/New_York")
sunrise <- times$sunrise
sunset <- times$sunset
return(ifelse(data > sunrise & data < sunset, 'dia', 'noite'))
}
dataset <- dataset %>%
mutate(TurnDay = getTurnDay(Human_timestamp))Algumas medidas estão multiplicadas por um fator de escala (padrão 10) e outras não estão no padrão de medida que gostariamos, para fins de análise posterior vamos convertê-las.
toCelsius <- function(f){
return((f - 32)/1.8)
}
toKg <- function(l){
return(l*0.4535925)
}
dataset <- dataset %>%
mutate(Ext_Temperature = Temperature/10,
Brood_Temp = toCelsius(Brood_Temp),
Hive_Temp = toCelsius(Hive_Temp),
Weight = toKg(Weight),
DewPoint = DewPoint/10) %>%
select(Season, TurnDay, Brood_Temp, Brood_Humidity, Hive_Temp,
Hive_Humidity, Weight, Ext_Temperature, DewPoint, WindDirection,
WindSpeed, Brood, Bees, Queen, Food, Stressors, Space)
summary(dataset)## Season TurnDay Brood_Temp Brood_Humidity
## Length:15490 Length:15490 Min. : 9.128 Min. :22.00
## Class :character Class :character 1st Qu.:29.347 1st Qu.:61.00
## Mode :character Mode :character Median :32.817 Median :66.00
## Mean :31.228 Mean :64.79
## 3rd Qu.:34.400 3rd Qu.:71.00
## Max. :43.961 Max. :87.00
## Hive_Temp Hive_Humidity Weight Ext_Temperature
## Min. :-17.78 Min. :19.00 Min. : 0.00 Min. : 0.00
## 1st Qu.: 27.83 1st Qu.:59.00 1st Qu.: 24.64 1st Qu.: 2.61
## Median : 32.50 Median :66.00 Median : 29.27 Median :19.40
## Mean : 30.62 Mean :64.39 Mean : 28.68 Mean :15.71
## 3rd Qu.: 34.53 3rd Qu.:71.00 3rd Qu.: 32.87 3rd Qu.:24.40
## Max. : 48.04 Max. :93.00 Max. :205.59 Max. :36.00
## DewPoint WindDirection WindSpeed Brood
## Min. :-6.70 Min. :-9999.0 Min. : 0.00 Min. :0.000
## 1st Qu.: 1.94 1st Qu.: 0.0 1st Qu.: 1.50 1st Qu.:1.000
## Median :15.00 Median : 100.0 Median :15.00 Median :1.000
## Mean :12.33 Mean : -528.3 Mean :17.46 Mean :0.846
## 3rd Qu.:21.10 3rd Qu.: 220.0 3rd Qu.:31.00 3rd Qu.:1.000
## Max. :27.00 Max. : 360.0 Max. :99.00 Max. :1.000
## Bees Queen Food Stressors
## Min. :0.0000 Min. :0.0000 Min. :0.0000 Min. :0.0000
## 1st Qu.:1.0000 1st Qu.:1.0000 1st Qu.:1.0000 1st Qu.:0.0000
## Median :1.0000 Median :1.0000 Median :1.0000 Median :0.0000
## Mean :0.9187 Mean :0.9022 Mean :0.9649 Mean :0.4362
## 3rd Qu.:1.0000 3rd Qu.:1.0000 3rd Qu.:1.0000 3rd Qu.:1.0000
## Max. :1.0000 Max. :1.0000 Max. :1.0000 Max. :1.0000
## Space
## Min. :0.0000
## 1st Qu.:0.0000
## Median :1.0000
## Mean :0.7104
## 3rd Qu.:1.0000
## Max. :1.0000Vamos deixar apenas os dados que possuam consistência… Substituindo por NA os que não são, para depois imputar
dataset <- dataset %>%
mutate(Brood_Temp = ifelse(between(Brood_Temp, -10, 40), Brood_Temp, NA),
Hive_Temp = ifelse(between(Hive_Temp, -10, 40), Hive_Temp, NA),
Brood_Humidity = ifelse(between(Brood_Humidity, 0, 100), Brood_Humidity, NA),
Hive_Humidity = ifelse(between(Hive_Humidity, 0, 100), Hive_Humidity, NA),
Weight = ifelse(between(Weight, 1, 130), Weight, NA),
Ext_Temperature = ifelse(between(Ext_Temperature, -10, 40), Ext_Temperature, NA),
DewPoint = ifelse(DewPoint == -999.9, NA, DewPoint),
WindDirection = ifelse(WindDirection == -9999, NA, WindDirection),
WindSpeed = ifelse(WindSpeed == -9999, NA, WindSpeed))
summary(dataset)## Season TurnDay Brood_Temp Brood_Humidity
## Length:15490 Length:15490 Min. : 9.128 Min. :22.00
## Class :character Class :character 1st Qu.:29.344 1st Qu.:61.00
## Mode :character Mode :character Median :32.817 Median :66.00
## Mean :31.223 Mean :64.79
## 3rd Qu.:34.400 3rd Qu.:71.00
## Max. :39.950 Max. :87.00
## NA's :8
## Hive_Temp Hive_Humidity Weight Ext_Temperature
## Min. : 6.861 Min. :19.00 Min. : 1.098 Min. : 0.00
## 1st Qu.:27.824 1st Qu.:59.00 1st Qu.: 26.045 1st Qu.: 2.61
## Median :32.483 Median :66.00 Median : 29.665 Median :19.40
## Mean :30.614 Mean :64.39 Mean : 30.722 Mean :15.71
## 3rd Qu.:34.528 3rd Qu.:71.00 3rd Qu.: 33.280 3rd Qu.:24.40
## Max. :39.928 Max. :93.00 Max. :129.936 Max. :36.00
## NA's :12 NA's :1130
## DewPoint WindDirection WindSpeed Brood
## Min. :-6.70 Min. : 0.0 Min. : 0.00 Min. :0.000
## 1st Qu.: 1.94 1st Qu.: 0.0 1st Qu.: 1.50 1st Qu.:1.000
## Median :15.00 Median :120.0 Median :15.00 Median :1.000
## Mean :12.33 Mean :125.3 Mean :17.46 Mean :0.846
## 3rd Qu.:21.10 3rd Qu.:230.0 3rd Qu.:31.00 3rd Qu.:1.000
## Max. :27.00 Max. :360.0 Max. :99.00 Max. :1.000
## NA's :1000
## Bees Queen Food Stressors
## Min. :0.0000 Min. :0.0000 Min. :0.0000 Min. :0.0000
## 1st Qu.:1.0000 1st Qu.:1.0000 1st Qu.:1.0000 1st Qu.:0.0000
## Median :1.0000 Median :1.0000 Median :1.0000 Median :0.0000
## Mean :0.9187 Mean :0.9022 Mean :0.9649 Mean :0.4362
## 3rd Qu.:1.0000 3rd Qu.:1.0000 3rd Qu.:1.0000 3rd Qu.:1.0000
## Max. :1.0000 Max. :1.0000 Max. :1.0000 Max. :1.0000
##
## Space
## Min. :0.0000
## 1st Qu.:0.0000
## Median :1.0000
## Mean :0.7104
## 3rd Qu.:1.0000
## Max. :1.0000
## Após especificar como NA os incosistentes, vamos imputa-los
df_imp <- mice::mice(dataset)##
## iter imp variable
## 1 1 Brood_Temp Hive_Temp Weight WindDirection
## 1 2 Brood_Temp Hive_Temp Weight WindDirection
## 1 3 Brood_Temp Hive_Temp Weight WindDirection
## 1 4 Brood_Temp Hive_Temp Weight WindDirection
## 1 5 Brood_Temp Hive_Temp Weight WindDirection
## 2 1 Brood_Temp Hive_Temp Weight WindDirection
## 2 2 Brood_Temp Hive_Temp Weight WindDirection
## 2 3 Brood_Temp Hive_Temp Weight WindDirection
## 2 4 Brood_Temp Hive_Temp Weight WindDirection
## 2 5 Brood_Temp Hive_Temp Weight WindDirection
## 3 1 Brood_Temp Hive_Temp Weight WindDirection
## 3 2 Brood_Temp Hive_Temp Weight WindDirection
## 3 3 Brood_Temp Hive_Temp Weight WindDirection
## 3 4 Brood_Temp Hive_Temp Weight WindDirection
## 3 5 Brood_Temp Hive_Temp Weight WindDirection
## 4 1 Brood_Temp Hive_Temp Weight WindDirection
## 4 2 Brood_Temp Hive_Temp Weight WindDirection
## 4 3 Brood_Temp Hive_Temp Weight WindDirection
## 4 4 Brood_Temp Hive_Temp Weight WindDirection
## 4 5 Brood_Temp Hive_Temp Weight WindDirection
## 5 1 Brood_Temp Hive_Temp Weight WindDirection
## 5 2 Brood_Temp Hive_Temp Weight WindDirection
## 5 3 Brood_Temp Hive_Temp Weight WindDirection
## 5 4 Brood_Temp Hive_Temp Weight WindDirection
## 5 5 Brood_Temp Hive_Temp Weight WindDirection## Warning: Number of logged events: 2dataset <- mice::complete(df_imp, 1)
summary(dataset)## Season TurnDay Brood_Temp Brood_Humidity
## Length:15490 Length:15490 Min. : 9.128 Min. :22.00
## Class :character Class :character 1st Qu.:29.347 1st Qu.:61.00
## Mode :character Mode :character Median :32.817 Median :66.00
## Mean :31.224 Mean :64.79
## 3rd Qu.:34.400 3rd Qu.:71.00
## Max. :39.950 Max. :87.00
## Hive_Temp Hive_Humidity Weight Ext_Temperature
## Min. : 6.861 Min. :19.00 Min. : 1.098 Min. : 0.00
## 1st Qu.:27.828 1st Qu.:59.00 1st Qu.: 25.397 1st Qu.: 2.61
## Median :32.483 Median :66.00 Median : 29.456 Median :19.40
## Mean :30.617 Mean :64.39 Mean : 30.456 Mean :15.71
## 3rd Qu.:34.528 3rd Qu.:71.00 3rd Qu.: 33.071 3rd Qu.:24.40
## Max. :39.928 Max. :93.00 Max. :129.936 Max. :36.00
## DewPoint WindDirection WindSpeed Brood
## Min. :-6.70 Min. : 0.0 Min. : 0.00 Min. :0.000
## 1st Qu.: 1.94 1st Qu.: 10.0 1st Qu.: 1.50 1st Qu.:1.000
## Median :15.00 Median :130.0 Median :15.00 Median :1.000
## Mean :12.33 Mean :126.2 Mean :17.46 Mean :0.846
## 3rd Qu.:21.10 3rd Qu.:230.0 3rd Qu.:31.00 3rd Qu.:1.000
## Max. :27.00 Max. :360.0 Max. :99.00 Max. :1.000
## Bees Queen Food Stressors
## Min. :0.0000 Min. :0.0000 Min. :0.0000 Min. :0.0000
## 1st Qu.:1.0000 1st Qu.:1.0000 1st Qu.:1.0000 1st Qu.:0.0000
## Median :1.0000 Median :1.0000 Median :1.0000 Median :0.0000
## Mean :0.9187 Mean :0.9022 Mean :0.9649 Mean :0.4362
## 3rd Qu.:1.0000 3rd Qu.:1.0000 3rd Qu.:1.0000 3rd Qu.:1.0000
## Max. :1.0000 Max. :1.0000 Max. :1.0000 Max. :1.0000
## Space
## Min. :0.0000
## 1st Qu.:0.0000
## Median :1.0000
## Mean :0.7104
## 3rd Qu.:1.0000
## Max. :1.0000Vamos transformar em Fator as colunas não numéricas
dataset <- dataset %>%
mutate(Season = as.factor(Season),
TurnDay = as_factor(TurnDay))Feito isso, temos o conjunto de dados final (Por enquanto… xD)
head(dataset, 10)## Season TurnDay Brood_Temp Brood_Humidity Hive_Temp Hive_Humidity
## 1 Primavera noite 29.97778 46 21.75556 33
## 2 Primavera noite 30.17778 48 21.68333 35
## 3 Primavera noite 30.78333 47 21.77778 35
## 4 Primavera noite 31.26111 51 21.85000 35
## 5 Primavera noite 33.17222 52 21.73889 39
## 6 Primavera noite 34.27222 50 21.75556 37
## 7 Primavera noite 34.44444 45 21.76667 37
## 8 Primavera dia 34.08889 39 21.26111 36
## 9 Primavera dia 34.13333 41 21.03333 36
## 10 Primavera dia 34.04444 37 20.78333 37
## Weight Ext_Temperature DewPoint WindDirection WindSpeed Brood Bees
## 1 38.60979 1.28 -0.94 190 3.6 1 1
## 2 38.61887 1.28 -0.78 180 3.6 1 1
## 3 38.65062 1.17 -0.56 200 4.6 1 1
## 4 38.62794 1.11 -0.22 200 3.1 1 1
## 5 38.86834 0.78 0.56 120 2.1 1 1
## 6 38.96813 0.78 0.56 150 2.6 1 1
## 7 38.90916 0.78 0.39 180 3.1 1 1
## 8 38.92731 0.83 0.33 210 3.6 1 1
## 9 38.86834 0.89 0.33 230 4.6 1 1
## 10 38.70958 0.89 0.39 210 5.1 1 1
## Queen Food Stressors Space
## 1 1 1 1 0
## 2 1 1 1 0
## 3 1 1 1 0
## 4 1 1 1 0
## 5 1 1 1 0
## 6 1 1 1 0
## 7 1 1 1 0
## 8 1 1 1 0
## 9 1 1 1 0
## 10 1 1 1 0Utilizaremos a função kmeansruns do pacote fpc, para verificar o melhor k baseado nas métrica Calinski-Harabasz. Antes, precisamos de um conjunto de dados totalmente NUMÉRICO, para isso utilizaremos a abordagem Casella de Referência.
dataset_cluster = dataset[, 1:11]
dataset_cluster = dummy_cols(dataset_cluster, select_columns = c('Season', 'TurnDay'),
remove_first_dummy = T)
dataset_cluster[, 3:11] = scale(dataset_cluster[, 3:11])
dataset_cluster = dataset_cluster[, 3:14]
km <- kmeansruns(dataset_cluster, krange = 2:10,criterion="ch", runs = 3,
iter.max = 60, critout = T)## 2 clusters 6859.209
## 3 clusters 4890.242
## 4 clusters 4291.702
## 5 clusters 4091.493
## 6 clusters 3638.465
## 7 clusters 3831.488
## 8 clusters 3099.589
## 9 clusters 3470.005
## 10 clusters 3239.475Função para a obtenção do Code, para um determinado agrupamento de fatores
getCode <- function(cols, comb){
soma = sum(cols)
comb = unlist(strsplit(comb, " "))
for(i in 1:3){
arr = as.integer(unlist(strsplit(comb[i], "")))
if(as.integer(soma) %in% arr){
code = i
}
}
return(code - 1)
}Para todos essas combinações rodaremos um algoritmo kNN, e escolheremos o que tiver maior acurácia
possible_comb <- c("1234 5 6", "1235 4 6", "1245 3 6", "1345 2 6", "2345 1 6", "1236 4 5",
"1246 3 5", "1346 2 5", "2346 1 5", "1256 3 4", "1356 2 4", "2356 1 4",
"1456 2 3", "2456 1 3", "3456 1 2", "1 23 456", "1 24 356", "1 25 346",
"1 26 345", "1 34 256", "1 35 246", "1 36 245", "1 45 236", "1 46 235",
"1 56 234", "2 13 456", "2 14 356", "2 15 346", "2 16 345", "2 34 156",
"2 35 146", "2 36 145", "2 45 136", "2 46 135", "2 56 134", "3 12 456",
"3 14 256", "3 15 246", "3 16 245", "3 24 156", "3 25 146", "3 26 145",
"3 45 126", "3 46 125", "3 56 124", "4 12 356", "4 13 256", "4 15 236",
"4 16 235", "4 23 156", "4 25 136", "4 26 135", "4 35 126", "4 36 125",
"4 56 123", "5 12 346", "5 13 246", "5 14 236", "5 16 234", "5 23 146",
"5 24 136", "5 26 134", "5 34 126", "5 36 124", "5 46 123", "6 12 345",
"6 13 245", "6 14 235", "6 15 234", "6 23 145", "6 24 135", "6 25 134",
"6 34 125", "6 35 124", "6 45 123")
arr_accuracy_knn <- vector()
#arr_accuracy_svm <- vector()
arr_accuracy_rf <- vector()
for(comb in possible_comb){
dataset_train <- bind_cols(dataset_cluster, dataset[,12:17])
dataset_train <- dataset_train %>%
mutate(index = 1:n()) %>%
group_by(index) %>%
mutate(Code = getCode(cbind(Brood, Bees, Queen, Food, Stressors, Space), comb))
dataset_train <- dataset_train %>%
ungroup() %>%
mutate(Code = as_factor(Code)) %>%
select(TurnDay_dia, "Season_Verão", Season_Outono, Brood_Temp, Brood_Humidity,
Hive_Temp, Hive_Humidity, Ext_Temperature, Weight, DewPoint, WindDirection,
WindSpeed, Code)
# apply algorthm knn
k <- 1:10
tunegrid <- expand.grid(.k = k)
knn = train(Code~., data = dataset_train, trControl = trainControl(method = "cv",
savePredictions='final'),
method = "knn", tuneGrid = tunegrid)
print(comb)
print(knn)
print(confusionMatrix(knn$pred$pred, knn$pred$obs))
arr_accuracy_knn <- append(arr_accuracy_knn, sum(diag(confusionMatrix(knn)$table))/100)
# apply algorthm svm
#tunegrid = expand.grid(.C = 1:3, .sigma = 1:4)
#svm_radial = train(Code~., data = dataset_train, method = "svmRadial",
# trControl = trainControl(method = "cv"),
# tuneGrid = tunegrid)
#arr_accuracy_svm <- append(arr_accuracy_svm, sum(diag(confusionMatrix(svm_radial)$table))/100)
# apply algorthm knn
rf = train(Code~., data = dataset_train, method = "ranger",
trControl = trainControl(method = "cv",
savePredictions = "final"))
print(rf)
print(confusionMatrix(rf$pred$pred, rf$pred$obs))
arr_accuracy_rf <- append(arr_accuracy_rf, sum(diag(confusionMatrix(rf)$table))/100)
print("----------------------------------------------------------------------------")
}## [1] "1234 5 6"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13940, 13941, 13940, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8957385 0.8325260
## 2 0.8693315 0.7902234
## 3 0.8756580 0.7996977
## 4 0.8635228 0.7799954
## 5 0.8620374 0.7774187
## 6 0.8532587 0.7629381
## 7 0.8524178 0.7615190
## 8 0.8452516 0.7494948
## 9 0.8400874 0.7408276
## 10 0.8384100 0.7375784
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 3972 318 91
## 1 276 7038 412
## 2 88 430 2865
##
## Overall Statistics
##
## Accuracy : 0.8957
## 95% CI : (0.8908, 0.9005)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8325
##
## Mcnemar's Test P-Value : 0.3333
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9161 0.9039 0.8507
## Specificity 0.9633 0.9107 0.9573
## Pos Pred Value 0.9066 0.9110 0.8469
## Neg Pred Value 0.9672 0.9037 0.9585
## Prevalence 0.2799 0.5026 0.2174
## Detection Rate 0.2564 0.4544 0.1850
## Detection Prevalence 0.2828 0.4988 0.2184
## Balanced Accuracy 0.9397 0.9073 0.9040
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13940, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9234331 0.8750737
## 2 extratrees 0.8774035 0.7952221
## 7 gini 0.9417679 0.9060494
## 7 extratrees 0.9397661 0.9022074
## 12 gini 0.9341501 0.8938265
## 12 extratrees 0.9442854 0.9098446
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 4067 97 30
## 1 203 7519 297
## 2 66 170 3041
##
## Overall Statistics
##
## Accuracy : 0.9443
## 95% CI : (0.9406, 0.9478)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9098
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9380 0.9657 0.9029
## Specificity 0.9886 0.9351 0.9805
## Pos Pred Value 0.9697 0.9376 0.9280
## Neg Pred Value 0.9762 0.9643 0.9732
## Prevalence 0.2799 0.5026 0.2174
## Detection Rate 0.2626 0.4854 0.1963
## Detection Prevalence 0.2708 0.5177 0.2116
## Balanced Accuracy 0.9633 0.9504 0.9417
## [1] "----------------------------------------------------------------------------"
## [1] "1235 4 6"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13940, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8907674 0.8096445
## 2 0.8644292 0.7640127
## 3 0.8730153 0.7775640
## 4 0.8588769 0.7527484
## 5 0.8569418 0.7487842
## 6 0.8505492 0.7368050
## 7 0.8471275 0.7298821
## 8 0.8382186 0.7134179
## 9 0.8367971 0.7100053
## 10 0.8320840 0.7010654
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 8256 321 430
## 1 323 2692 88
## 2 444 86 2850
##
## Overall Statistics
##
## Accuracy : 0.8908
## 95% CI : (0.8858, 0.8956)
## No Information Rate : 0.5825
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8096
##
## Mcnemar's Test P-Value : 0.9685
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9150 0.8687 0.8462
## Specificity 0.8839 0.9668 0.9563
## Pos Pred Value 0.9166 0.8675 0.8432
## Neg Pred Value 0.8817 0.9671 0.9572
## Prevalence 0.5825 0.2001 0.2174
## Detection Rate 0.5330 0.1738 0.1840
## Detection Prevalence 0.5815 0.2003 0.2182
## Balanced Accuracy 0.8994 0.9177 0.9012
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13940, 13943, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9183353 0.8532677
## 2 extratrees 0.8635244 0.7431624
## 7 gini 0.9357654 0.8866569
## 7 extratrees 0.9360237 0.8861997
## 12 gini 0.9290498 0.8748801
## 12 extratrees 0.9398974 0.8935313
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 8779 294 310
## 1 72 2745 23
## 2 172 60 3035
##
## Overall Statistics
##
## Accuracy : 0.9399
## 95% CI : (0.936, 0.9436)
## No Information Rate : 0.5825
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8935
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9730 0.8858 0.9011
## Specificity 0.9066 0.9923 0.9809
## Pos Pred Value 0.9356 0.9665 0.9290
## Neg Pred Value 0.9600 0.9720 0.9728
## Prevalence 0.5825 0.2001 0.2174
## Detection Rate 0.5668 0.1772 0.1959
## Detection Prevalence 0.6057 0.1833 0.2109
## Balanced Accuracy 0.9398 0.9391 0.9410
## [1] "----------------------------------------------------------------------------"
## [1] "1245 3 6"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9295029 0.8102545
## 2 0.9100065 0.7586932
## 3 0.9167205 0.7733324
## 4 0.9050355 0.7421841
## 5 0.9066495 0.7442475
## 6 0.9001291 0.7255138
## 7 0.8988380 0.7200340
## 8 0.8937379 0.7048486
## 9 0.8921240 0.6988553
## 10 0.8867011 0.6828646
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 11252 36 488
## 1 44 266 0
## 2 524 0 2880
##
## Overall Statistics
##
## Accuracy : 0.9295
## 95% CI : (0.9254, 0.9335)
## No Information Rate : 0.7631
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8103
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9519 0.88079 0.8551
## Specificity 0.8572 0.99710 0.9568
## Pos Pred Value 0.9555 0.85806 0.8461
## Neg Pred Value 0.8471 0.99763 0.9596
## Prevalence 0.7631 0.01950 0.2174
## Detection Rate 0.7264 0.01717 0.1859
## Detection Prevalence 0.7602 0.02001 0.2198
## Balanced Accuracy 0.9046 0.93895 0.9059
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9375071 0.8196579
## 2 extratrees 0.8941901 0.6668151
## 7 gini 0.9540347 0.8725007
## 7 extratrees 0.9521618 0.8654677
## 12 gini 0.9510006 0.8646071
## 12 extratrees 0.9549381 0.8743969
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 11617 63 432
## 1 6 239 0
## 2 197 0 2936
##
## Overall Statistics
##
## Accuracy : 0.9549
## 95% CI : (0.9516, 0.9582)
## No Information Rate : 0.7631
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8745
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9828 0.79139 0.8717
## Specificity 0.8651 0.99960 0.9837
## Pos Pred Value 0.9591 0.97551 0.9371
## Neg Pred Value 0.9399 0.99587 0.9650
## Prevalence 0.7631 0.01950 0.2174
## Detection Rate 0.7500 0.01543 0.1895
## Detection Prevalence 0.7819 0.01582 0.2023
## Balanced Accuracy 0.9240 0.89550 0.9277
## [1] "----------------------------------------------------------------------------"
## [1] "1345 2 6"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13940, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9224010 0.8077579
## 2 0.9020647 0.7589109
## 3 0.9103929 0.7763510
## 4 0.8972227 0.7433670
## 5 0.8996766 0.7466887
## 6 0.8910900 0.7248251
## 7 0.8874750 0.7134595
## 8 0.8843772 0.7055136
## 9 0.8802447 0.6928797
## 10 0.8760484 0.6812457
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 10859 66 510
## 1 91 574 3
## 2 530 2 2855
##
## Overall Statistics
##
## Accuracy : 0.9224
## 95% CI : (0.9181, 0.9266)
## No Information Rate : 0.7411
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8077
##
## Mcnemar's Test P-Value : 0.2065
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9459 0.89408 0.8477
## Specificity 0.8564 0.99367 0.9561
## Pos Pred Value 0.9496 0.85928 0.8429
## Neg Pred Value 0.8469 0.99541 0.9576
## Prevalence 0.7411 0.04145 0.2174
## Detection Rate 0.7010 0.03706 0.1843
## Detection Prevalence 0.7382 0.04312 0.2187
## Balanced Accuracy 0.9011 0.94388 0.9019
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13941, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9367977 0.8334365
## 2 extratrees 0.8917382 0.6902818
## 7 gini 0.9534530 0.8818254
## 7 extratrees 0.9525496 0.8777207
## 12 gini 0.9495791 0.8724179
## 12 extratrees 0.9564878 0.8886099
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 11294 62 426
## 1 11 580 0
## 2 175 0 2942
##
## Overall Statistics
##
## Accuracy : 0.9565
## 95% CI : (0.9532, 0.9596)
## No Information Rate : 0.7411
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8887
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9838 0.90343 0.8735
## Specificity 0.8783 0.99926 0.9856
## Pos Pred Value 0.9586 0.98139 0.9439
## Neg Pred Value 0.9498 0.99584 0.9656
## Prevalence 0.7411 0.04145 0.2174
## Detection Rate 0.7291 0.03744 0.1899
## Detection Prevalence 0.7606 0.03815 0.2012
## Balanced Accuracy 0.9311 0.95134 0.9295
## [1] "----------------------------------------------------------------------------"
## [1] "2345 1 6"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13942, 13940, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9281479 0.8058560
## 2 0.9086513 0.7532331
## 3 0.9142021 0.7657617
## 4 0.9061331 0.7428669
## 5 0.9058108 0.7395553
## 6 0.8958684 0.7114458
## 7 0.8961270 0.7090499
## 8 0.8912861 0.6947532
## 9 0.8883174 0.6843687
## 10 0.8823777 0.6675162
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 11257 35 505
## 1 35 258 1
## 2 537 0 2862
##
## Overall Statistics
##
## Accuracy : 0.9281
## 95% CI : (0.924, 0.9322)
## No Information Rate : 0.7637
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.806
##
## Mcnemar's Test P-Value : 0.576
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9516 0.88055 0.8498
## Specificity 0.8525 0.99763 0.9557
## Pos Pred Value 0.9542 0.87755 0.8420
## Neg Pred Value 0.8451 0.99770 0.9582
## Prevalence 0.7637 0.01892 0.2174
## Detection Rate 0.7267 0.01666 0.1848
## Detection Prevalence 0.7616 0.01898 0.2194
## Balanced Accuracy 0.9021 0.93909 0.9027
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13942, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9408654 0.8301053
## 2 extratrees 0.8991606 0.6841064
## 7 gini 0.9573273 0.8824050
## 7 extratrees 0.9543574 0.8717687
## 12 gini 0.9533246 0.8718571
## 12 extratrees 0.9582307 0.8837862
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 11638 36 420
## 1 6 257 0
## 2 185 0 2948
##
## Overall Statistics
##
## Accuracy : 0.9582
## 95% CI : (0.955, 0.9613)
## No Information Rate : 0.7637
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8838
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9839 0.87713 0.8753
## Specificity 0.8754 0.99961 0.9847
## Pos Pred Value 0.9623 0.97719 0.9410
## Neg Pred Value 0.9438 0.99764 0.9660
## Prevalence 0.7637 0.01892 0.2174
## Detection Rate 0.7513 0.01659 0.1903
## Detection Prevalence 0.7808 0.01698 0.2023
## Balanced Accuracy 0.9296 0.93837 0.9300
## [1] "----------------------------------------------------------------------------"
## [1] "1236 4 5"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13940, 13942, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8885094 0.8199568
## 2 0.8587478 0.7720916
## 3 0.8664948 0.7837336
## 4 0.8533901 0.7621033
## 5 0.8519684 0.7595906
## 6 0.8432534 0.7450869
## 7 0.8404133 0.7401393
## 8 0.8327295 0.7271943
## 9 0.8314386 0.7246466
## 10 0.8267253 0.7164787
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 4032 154 459
## 1 134 2678 274
## 2 439 267 7053
##
## Overall Statistics
##
## Accuracy : 0.8885
## 95% CI : (0.8834, 0.8934)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8199
##
## Mcnemar's Test P-Value : 0.5881
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8756 0.8641 0.9059
## Specificity 0.9437 0.9671 0.9084
## Pos Pred Value 0.8680 0.8678 0.9090
## Neg Pred Value 0.9472 0.9661 0.9052
## Prevalence 0.2973 0.2001 0.5026
## Detection Rate 0.2603 0.1729 0.4553
## Detection Prevalence 0.2999 0.1992 0.5009
## Balanced Accuracy 0.9096 0.9156 0.9071
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13942, 13940, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9205285 0.8694852
## 2 extratrees 0.8724985 0.7859486
## 7 gini 0.9398968 0.9022353
## 7 extratrees 0.9369912 0.8971195
## 12 gini 0.9371195 0.8977688
## 12 extratrees 0.9420910 0.9057052
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 4245 79 183
## 1 56 2806 61
## 2 304 214 7542
##
## Overall Statistics
##
## Accuracy : 0.9421
## 95% CI : (0.9383, 0.9457)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9057
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9218 0.9055 0.9687
## Specificity 0.9759 0.9906 0.9328
## Pos Pred Value 0.9419 0.9600 0.9357
## Neg Pred Value 0.9672 0.9767 0.9672
## Prevalence 0.2973 0.2001 0.5026
## Detection Rate 0.2740 0.1811 0.4869
## Detection Prevalence 0.2910 0.1887 0.5203
## Balanced Accuracy 0.9489 0.9480 0.9507
## [1] "----------------------------------------------------------------------------"
## [1] "1246 3 5"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13940, 13941, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9056169 0.8182065
## 2 0.8805678 0.7699273
## 3 0.8866365 0.7814649
## 4 0.8735322 0.7561359
## 5 0.8752112 0.7592595
## 6 0.8666251 0.7427102
## 7 0.8646885 0.7390009
## 8 0.8574580 0.7249833
## 9 0.8549396 0.7199636
## 10 0.8493874 0.7092256
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 6695 29 707
## 1 37 268 14
## 2 670 5 7065
##
## Overall Statistics
##
## Accuracy : 0.9056
## 95% CI : (0.9009, 0.9102)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8182
##
## Mcnemar's Test P-Value : 0.1011
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9045 0.88742 0.9074
## Specificity 0.9090 0.99664 0.9124
## Pos Pred Value 0.9010 0.84013 0.9128
## Neg Pred Value 0.9123 0.99776 0.9070
## Prevalence 0.4779 0.01950 0.5026
## Detection Rate 0.4322 0.01730 0.4561
## Detection Prevalence 0.4797 0.02059 0.4997
## Balanced Accuracy 0.9067 0.94203 0.9099
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13940, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9338286 0.8718275
## 2 extratrees 0.8941252 0.7941274
## 7 gini 0.9446098 0.8927578
## 7 extratrees 0.9479021 0.8992682
## 12 gini 0.9402845 0.8844366
## 12 extratrees 0.9508714 0.9050414
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 6992 39 310
## 1 2 262 1
## 2 408 1 7475
##
## Overall Statistics
##
## Accuracy : 0.9509
## 95% CI : (0.9473, 0.9542)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.905
##
## Mcnemar's Test P-Value : 3.897e-10
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9446 0.86755 0.9601
## Specificity 0.9568 0.99980 0.9469
## Pos Pred Value 0.9525 0.98868 0.9481
## Neg Pred Value 0.9497 0.99737 0.9591
## Prevalence 0.4779 0.01950 0.5026
## Detection Rate 0.4514 0.01691 0.4826
## Detection Prevalence 0.4739 0.01711 0.5090
## Balanced Accuracy 0.9507 0.93368 0.9535
## [1] "----------------------------------------------------------------------------"
## [1] "1346 2 5"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8975476 0.8097296
## 2 0.8737251 0.7657420
## 3 0.8779857 0.7732198
## 4 0.8631377 0.7457131
## 5 0.8659132 0.7507500
## 6 0.8529368 0.7264237
## 7 0.8517749 0.7244261
## 8 0.8456419 0.7127910
## 9 0.8449316 0.7112976
## 10 0.8391862 0.7004602
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 6294 61 729
## 1 73 571 19
## 2 695 10 7038
##
## Overall Statistics
##
## Accuracy : 0.8975
## 95% CI : (0.8927, 0.9023)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8097
##
## Mcnemar's Test P-Value : 0.1968
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8912 0.88941 0.9039
## Specificity 0.9063 0.99380 0.9085
## Pos Pred Value 0.8885 0.86124 0.9090
## Neg Pred Value 0.9086 0.99521 0.9034
## Prevalence 0.4559 0.04145 0.5026
## Detection Rate 0.4063 0.03686 0.4544
## Detection Prevalence 0.4573 0.04280 0.4999
## Balanced Accuracy 0.8988 0.94161 0.9062
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9326011 0.8741421
## 2 extratrees 0.8923826 0.7974472
## 7 gini 0.9453837 0.8983034
## 7 extratrees 0.9475142 0.9021948
## 12 gini 0.9406712 0.8895045
## 12 extratrees 0.9498386 0.9065776
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 6626 39 297
## 1 18 602 4
## 2 418 1 7485
##
## Overall Statistics
##
## Accuracy : 0.9498
## 95% CI : (0.9463, 0.9532)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9066
##
## Mcnemar's Test P-Value : 1.371e-06
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9383 0.93769 0.9613
## Specificity 0.9601 0.99852 0.9456
## Pos Pred Value 0.9517 0.96474 0.9470
## Neg Pred Value 0.9489 0.99731 0.9603
## Prevalence 0.4559 0.04145 0.5026
## Detection Rate 0.4278 0.03886 0.4832
## Detection Prevalence 0.4495 0.04028 0.5103
## Balanced Accuracy 0.9492 0.96811 0.9535
## [1] "----------------------------------------------------------------------------"
## [1] "2346 1 5"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13942, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9025170 0.8118299
## 2 0.8775989 0.7637036
## 3 0.8842487 0.7762784
## 4 0.8705619 0.7497348
## 5 0.8715949 0.7515496
## 6 0.8597161 0.7282875
## 7 0.8613305 0.7312631
## 8 0.8536473 0.7162846
## 9 0.8514524 0.7120105
## 10 0.8467402 0.7026252
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 6685 37 742
## 1 35 253 2
## 2 691 3 7042
##
## Overall Statistics
##
## Accuracy : 0.9025
## 95% CI : (0.8977, 0.9071)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8118
##
## Mcnemar's Test P-Value : 0.5579
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9020 0.86348 0.9044
## Specificity 0.9036 0.99757 0.9099
## Pos Pred Value 0.8956 0.87241 0.9103
## Neg Pred Value 0.9095 0.99737 0.9040
## Prevalence 0.4784 0.01892 0.5026
## Detection Rate 0.4316 0.01633 0.4546
## Detection Prevalence 0.4819 0.01872 0.4994
## Balanced Accuracy 0.9028 0.93052 0.9072
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13941, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9341519 0.8723794
## 2 extratrees 0.8948350 0.7952266
## 7 gini 0.9487423 0.9008711
## 7 extratrees 0.9476440 0.8986914
## 12 gini 0.9446756 0.8929931
## 12 extratrees 0.9508076 0.9048345
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 6986 39 298
## 1 6 254 0
## 2 419 0 7488
##
## Overall Statistics
##
## Accuracy : 0.9508
## 95% CI : (0.9473, 0.9542)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9048
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9427 0.86689 0.9617
## Specificity 0.9583 0.99961 0.9456
## Pos Pred Value 0.9540 0.97692 0.9470
## Neg Pred Value 0.9480 0.99744 0.9607
## Prevalence 0.4784 0.01892 0.5026
## Detection Rate 0.4510 0.01640 0.4834
## Detection Prevalence 0.4728 0.01679 0.5105
## Balanced Accuracy 0.9505 0.93325 0.9537
## [1] "----------------------------------------------------------------------------"
## [1] "1256 3 4"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13941, 13940, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9445449 0.8418987
## 2 0.9322146 0.8066319
## 3 0.9352484 0.8111670
## 4 0.9289220 0.7925299
## 5 0.9306647 0.7955479
## 6 0.9245312 0.7763390
## 7 0.9244669 0.7746168
## 8 0.9211744 0.7633260
## 9 0.9209163 0.7610427
## 10 0.9176882 0.7502188
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 11674 25 387
## 1 27 264 19
## 2 388 13 2693
##
## Overall Statistics
##
## Accuracy : 0.9445
## 95% CI : (0.9408, 0.9481)
## No Information Rate : 0.7804
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8419
##
## Mcnemar's Test P-Value : 0.7522
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9657 0.87417 0.8690
## Specificity 0.8789 0.99697 0.9676
## Pos Pred Value 0.9659 0.85161 0.8704
## Neg Pred Value 0.8781 0.99750 0.9672
## Prevalence 0.7804 0.01950 0.2001
## Detection Rate 0.7536 0.01704 0.1739
## Detection Prevalence 0.7802 0.02001 0.1997
## Balanced Accuracy 0.9223 0.93557 0.9183
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9488701 0.8427540
## 2 extratrees 0.9218204 0.7450876
## 7 gini 0.9632663 0.8908103
## 7 extratrees 0.9635253 0.8909613
## 12 gini 0.9604906 0.8832798
## 12 extratrees 0.9666238 0.9008160
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 12016 34 395
## 1 3 256 3
## 2 70 12 2701
##
## Overall Statistics
##
## Accuracy : 0.9666
## 95% CI : (0.9637, 0.9694)
## No Information Rate : 0.7804
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9009
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9940 0.84768 0.8716
## Specificity 0.8739 0.99960 0.9934
## Pos Pred Value 0.9655 0.97710 0.9705
## Neg Pred Value 0.9760 0.99698 0.9687
## Prevalence 0.7804 0.01950 0.2001
## Detection Rate 0.7757 0.01653 0.1744
## Detection Prevalence 0.8034 0.01691 0.1797
## Balanced Accuracy 0.9339 0.92364 0.9325
## [1] "----------------------------------------------------------------------------"
## [1] "1356 2 4"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13942, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9416385 0.8476907
## 2 0.9251761 0.8047578
## 3 0.9309215 0.8164223
## 4 0.9255637 0.8020192
## 5 0.9250475 0.7977707
## 6 0.9200761 0.7846972
## 7 0.9182675 0.7784275
## 8 0.9140069 0.7657837
## 9 0.9136850 0.7631706
## 10 0.9109747 0.7553266
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 11325 42 379
## 1 52 573 32
## 2 372 27 2688
##
## Overall Statistics
##
## Accuracy : 0.9416
## 95% CI : (0.9378, 0.9453)
## No Information Rate : 0.7585
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8477
##
## Mcnemar's Test P-Value : 0.6701
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9639 0.89252 0.8674
## Specificity 0.8875 0.99434 0.9678
## Pos Pred Value 0.9642 0.87215 0.8707
## Neg Pred Value 0.8868 0.99535 0.9669
## Prevalence 0.7585 0.04145 0.2001
## Detection Rate 0.7311 0.03699 0.1735
## Detection Prevalence 0.7583 0.04241 0.1993
## Balanced Accuracy 0.9257 0.94343 0.9176
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9502905 0.8620464
## 2 extratrees 0.9225307 0.7728497
## 7 gini 0.9642350 0.9036018
## 7 extratrees 0.9642350 0.9031105
## 12 gini 0.9603615 0.8939358
## 12 extratrees 0.9667527 0.9103340
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 11669 19 389
## 1 11 604 8
## 2 69 19 2702
##
## Overall Statistics
##
## Accuracy : 0.9668
## 95% CI : (0.9638, 0.9695)
## No Information Rate : 0.7585
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9104
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9932 0.94081 0.8719
## Specificity 0.8909 0.99872 0.9929
## Pos Pred Value 0.9662 0.96950 0.9685
## Neg Pred Value 0.9766 0.99744 0.9687
## Prevalence 0.7585 0.04145 0.2001
## Detection Rate 0.7533 0.03899 0.1744
## Detection Prevalence 0.7797 0.04022 0.1801
## Balanced Accuracy 0.9421 0.96977 0.9324
## [1] "----------------------------------------------------------------------------"
## [1] "2356 1 4"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13941, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9445445 0.8410832
## 2 0.9302780 0.8000518
## 3 0.9335701 0.8055932
## 4 0.9277590 0.7874548
## 5 0.9278236 0.7849285
## 6 0.9235625 0.7712647
## 7 0.9220780 0.7641740
## 8 0.9191088 0.7546336
## 9 0.9176233 0.7479719
## 10 0.9148473 0.7387510
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 11680 32 401
## 1 30 257 4
## 2 388 4 2694
##
## Overall Statistics
##
## Accuracy : 0.9445
## 95% CI : (0.9408, 0.9481)
## No Information Rate : 0.781
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8411
##
## Mcnemar's Test P-Value : 0.964
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9654 0.87713 0.8693
## Specificity 0.8723 0.99776 0.9684
## Pos Pred Value 0.9643 0.88316 0.8730
## Neg Pred Value 0.8762 0.99763 0.9673
## Prevalence 0.7810 0.01892 0.2001
## Detection Rate 0.7540 0.01659 0.1739
## Detection Prevalence 0.7820 0.01879 0.1992
## Balanced Accuracy 0.9189 0.93745 0.9188
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13942, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9480299 0.8390161
## 2 extratrees 0.9214318 0.7414374
## 7 gini 0.9630727 0.8899991
## 7 extratrees 0.9625554 0.8875365
## 12 gini 0.9608120 0.8840464
## 12 extratrees 0.9655899 0.8975417
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 12009 31 411
## 1 8 260 0
## 2 81 2 2688
##
## Overall Statistics
##
## Accuracy : 0.9656
## 95% CI : (0.9626, 0.9684)
## No Information Rate : 0.781
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8976
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9926 0.88737 0.8674
## Specificity 0.8697 0.99947 0.9933
## Pos Pred Value 0.9645 0.97015 0.9700
## Neg Pred Value 0.9707 0.99783 0.9677
## Prevalence 0.7810 0.01892 0.2001
## Detection Rate 0.7753 0.01679 0.1735
## Detection Prevalence 0.8038 0.01730 0.1789
## Balanced Accuracy 0.9312 0.94342 0.9303
## [1] "----------------------------------------------------------------------------"
## [1] "1456 2 3"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13941, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9854742 0.8776215
## 2 0.9805031 0.8359573
## 3 0.9815364 0.8418572
## 4 0.9791483 0.8218837
## 5 0.9779212 0.8086813
## 6 0.9766956 0.7973063
## 7 0.9763725 0.7935745
## 8 0.9748235 0.7775643
## 9 0.9734677 0.7666213
## 10 0.9728864 0.7577610
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 14421 59 23
## 1 83 577 12
## 2 42 6 267
##
## Overall Statistics
##
## Accuracy : 0.9855
## 95% CI : (0.9835, 0.9873)
## No Information Rate : 0.9391
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8775
##
## Mcnemar's Test P-Value : 0.008845
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9914 0.89875 0.88411
## Specificity 0.9131 0.99360 0.99684
## Pos Pred Value 0.9943 0.85863 0.84762
## Neg Pred Value 0.8734 0.99561 0.99769
## Prevalence 0.9391 0.04145 0.01950
## Detection Rate 0.9310 0.03725 0.01724
## Detection Prevalence 0.9363 0.04338 0.02034
## Balanced Accuracy 0.9523 0.94618 0.94047
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13940, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9841837 0.8469272
## 2 extratrees 0.9758558 0.7446382
## 7 gini 0.9907041 0.9160258
## 7 extratrees 0.9907038 0.9153437
## 12 gini 0.9902522 0.9130562
## 12 extratrees 0.9908977 0.9176338
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 14520 61 49
## 1 21 577 1
## 2 5 4 252
##
## Overall Statistics
##
## Accuracy : 0.9909
## 95% CI : (0.9893, 0.9923)
## No Information Rate : 0.9391
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9181
##
## Mcnemar's Test P-Value : 2.371e-12
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9982 0.89875 0.83444
## Specificity 0.8835 0.99852 0.99941
## Pos Pred Value 0.9925 0.96327 0.96552
## Neg Pred Value 0.9698 0.99563 0.99672
## Prevalence 0.9391 0.04145 0.01950
## Detection Rate 0.9374 0.03725 0.01627
## Detection Prevalence 0.9445 0.03867 0.01685
## Balanced Accuracy 0.9408 0.94864 0.91692
## [1] "----------------------------------------------------------------------------"
## [1] "2456 1 3"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13942, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9901872 0.8696145
## 2 0.9885734 0.8476145
## 3 0.9881863 0.8389193
## 4 0.9870241 0.8201332
## 5 0.9868951 0.8149409
## 6 0.9859913 0.8015183
## 7 0.9854750 0.7901044
## 8 0.9852814 0.7865825
## 9 0.9852812 0.7858801
## 10 0.9839255 0.7649105
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 14820 36 28
## 1 33 250 6
## 2 42 7 268
##
## Overall Statistics
##
## Accuracy : 0.9902
## 95% CI : (0.9885, 0.9917)
## No Information Rate : 0.9616
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8696
##
## Mcnemar's Test P-Value : 0.3905
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9950 0.85324 0.88742
## Specificity 0.8924 0.99743 0.99677
## Pos Pred Value 0.9957 0.86505 0.84543
## Neg Pred Value 0.8762 0.99717 0.99776
## Prevalence 0.9616 0.01892 0.01950
## Detection Rate 0.9567 0.01614 0.01730
## Detection Prevalence 0.9609 0.01866 0.02046
## Balanced Accuracy 0.9437 0.92534 0.94210
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13942, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9894120 0.8379795
## 2 extratrees 0.9835376 0.7212045
## 7 gini 0.9934795 0.9062804
## 7 extratrees 0.9930275 0.8985995
## 12 gini 0.9928337 0.8978661
## 12 extratrees 0.9936730 0.9091508
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 14886 40 45
## 1 6 252 3
## 2 3 1 254
##
## Overall Statistics
##
## Accuracy : 0.9937
## 95% CI : (0.9923, 0.9949)
## No Information Rate : 0.9616
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9096
##
## Mcnemar's Test P-Value : 1.424e-13
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9994 0.86007 0.84106
## Specificity 0.8571 0.99941 0.99974
## Pos Pred Value 0.9943 0.96552 0.98450
## Neg Pred Value 0.9827 0.99731 0.99685
## Prevalence 0.9616 0.01892 0.01950
## Detection Rate 0.9610 0.01627 0.01640
## Detection Prevalence 0.9665 0.01685 0.01666
## Balanced Accuracy 0.9283 0.92974 0.92040
## [1] "----------------------------------------------------------------------------"
## [1] "3456 1 2"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13942, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9883150 0.8993246
## 2 0.9828926 0.8530462
## 3 0.9840542 0.8596685
## 4 0.9823754 0.8431761
## 5 0.9796640 0.8178259
## 6 0.9785664 0.8064065
## 7 0.9783727 0.8021538
## 8 0.9772751 0.7924700
## 9 0.9766946 0.7849215
## 10 0.9762426 0.7792604
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 14479 12 44
## 1 16 254 22
## 2 60 27 576
##
## Overall Statistics
##
## Accuracy : 0.9883
## 95% CI : (0.9865, 0.9899)
## No Information Rate : 0.9396
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8994
##
## Mcnemar's Test P-Value : 0.3152
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9948 0.86689 0.89720
## Specificity 0.9401 0.99750 0.99414
## Pos Pred Value 0.9961 0.86986 0.86878
## Neg Pred Value 0.9204 0.99743 0.99555
## Prevalence 0.9396 0.01892 0.04145
## Detection Rate 0.9347 0.01640 0.03719
## Detection Prevalence 0.9383 0.01885 0.04280
## Balanced Accuracy 0.9674 0.93220 0.94567
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13940, 13939, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9896067 0.9031705
## 2 extratrees 0.9804394 0.8012781
## 7 gini 0.9916078 0.9243310
## 7 extratrees 0.9934157 0.9409461
## 12 gini 0.9908332 0.9178016
## 12 extratrees 0.9936739 0.9434571
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 14540 8 53
## 1 6 268 5
## 2 9 17 584
##
## Overall Statistics
##
## Accuracy : 0.9937
## 95% CI : (0.9923, 0.9949)
## No Information Rate : 0.9396
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9437
##
## Mcnemar's Test P-Value : 2.749e-08
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9990 0.91468 0.90966
## Specificity 0.9348 0.99928 0.99825
## Pos Pred Value 0.9958 0.96057 0.95738
## Neg Pred Value 0.9831 0.99836 0.99610
## Prevalence 0.9396 0.01892 0.04145
## Detection Rate 0.9387 0.01730 0.03770
## Detection Prevalence 0.9426 0.01801 0.03938
## Balanced Accuracy 0.9669 0.95698 0.95395
## [1] "----------------------------------------------------------------------------"
## [1] "1 23 456"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13942, 13941, 13939, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9848933 0.9000915
## 2 0.9804391 0.8700037
## 3 0.9811492 0.8739801
## 4 0.9787606 0.8569639
## 5 0.9775341 0.8474167
## 6 0.9761776 0.8372975
## 7 0.9753382 0.8296537
## 8 0.9744986 0.8236564
## 9 0.9734012 0.8151805
## 10 0.9719814 0.8047664
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 255 30 7
## 1 30 853 98
## 2 8 61 14148
##
## Overall Statistics
##
## Accuracy : 0.9849
## 95% CI : (0.9828, 0.9868)
## No Information Rate : 0.9201
## P-Value [Acc > NIR] : < 2e-16
##
## Kappa : 0.9001
##
## Mcnemar's Test P-Value : 0.03391
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87031 0.90360 0.9926
## Specificity 0.99757 0.99120 0.9442
## Pos Pred Value 0.87329 0.86952 0.9951
## Neg Pred Value 0.99750 0.99373 0.9175
## Prevalence 0.01892 0.06094 0.9201
## Detection Rate 0.01646 0.05507 0.9134
## Detection Prevalence 0.01885 0.06333 0.9178
## Balanced Accuracy 0.93394 0.94740 0.9684
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13942, 13942, 13940, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9863130 0.9027355
## 2 extratrees 0.9775981 0.8322904
## 7 gini 0.9910260 0.9382632
## 7 extratrees 0.9903161 0.9334238
## 12 gini 0.9903806 0.9344074
## 12 extratrees 0.9912844 0.9403638
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 257 9 1
## 1 29 868 22
## 2 7 67 14230
##
## Overall Statistics
##
## Accuracy : 0.9913
## 95% CI : (0.9897, 0.9927)
## No Information Rate : 0.9201
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9405
##
## Mcnemar's Test P-Value : 3.148e-08
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87713 0.91949 0.9984
## Specificity 0.99934 0.99649 0.9402
## Pos Pred Value 0.96255 0.94450 0.9948
## Neg Pred Value 0.99764 0.99478 0.9806
## Prevalence 0.01892 0.06094 0.9201
## Detection Rate 0.01659 0.05604 0.9187
## Detection Prevalence 0.01724 0.05933 0.9234
## Balanced Accuracy 0.93824 0.95799 0.9693
## [1] "----------------------------------------------------------------------------"
## [1] "1 24 356"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13940, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9439638 0.8583334
## 2 0.9292439 0.8205145
## 3 0.9358291 0.8350418
## 4 0.9280176 0.8148172
## 5 0.9286629 0.8145842
## 6 0.9253062 0.8047834
## 7 0.9233698 0.7992241
## 8 0.9201417 0.7903468
## 9 0.9198837 0.7886102
## 10 0.9173659 0.7816877
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 249 28 9
## 1 35 3339 413
## 2 9 374 11034
##
## Overall Statistics
##
## Accuracy : 0.944
## 95% CI : (0.9402, 0.9475)
## No Information Rate : 0.7396
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8583
##
## Mcnemar's Test P-Value : 0.4385
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.84983 0.8925 0.9632
## Specificity 0.99757 0.9619 0.9051
## Pos Pred Value 0.87063 0.8817 0.9665
## Neg Pred Value 0.99711 0.9656 0.8964
## Prevalence 0.01892 0.2415 0.7396
## Detection Rate 0.01607 0.2156 0.7123
## Detection Prevalence 0.01846 0.2445 0.7371
## Balanced Accuracy 0.92370 0.9272 0.9341
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13942, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9535183 0.8766434
## 2 extratrees 0.9315687 0.8124564
## 7 gini 0.9663009 0.9126361
## 7 extratrees 0.9648162 0.9082348
## 12 gini 0.9609428 0.8990826
## 12 extratrees 0.9679795 0.9168397
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 256 5 4
## 1 33 3375 89
## 2 4 361 11363
##
## Overall Statistics
##
## Accuracy : 0.968
## 95% CI : (0.9651, 0.9707)
## No Information Rate : 0.7396
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9169
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87372 0.9022 0.9919
## Specificity 0.99941 0.9896 0.9095
## Pos Pred Value 0.96604 0.9651 0.9689
## Neg Pred Value 0.99757 0.9695 0.9753
## Prevalence 0.01892 0.2415 0.7396
## Detection Rate 0.01653 0.2179 0.7336
## Detection Prevalence 0.01711 0.2258 0.7571
## Balanced Accuracy 0.93656 0.9459 0.9507
## [1] "----------------------------------------------------------------------------"
## [1] "1 25 346"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8998703 0.8047451
## 2 0.8704964 0.7474467
## 3 0.8785662 0.7627410
## 4 0.8659119 0.7379551
## 5 0.8664287 0.7386648
## 6 0.8557124 0.7174092
## 7 0.8555186 0.7167282
## 8 0.8448668 0.6955350
## 9 0.8460929 0.6978347
## 10 0.8409284 0.6876961
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 253 28 13
## 1 28 7653 723
## 2 12 747 6033
##
## Overall Statistics
##
## Accuracy : 0.8999
## 95% CI : (0.895, 0.9046)
## No Information Rate : 0.5441
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8048
##
## Mcnemar's Test P-Value : 0.9336
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.86348 0.9080 0.8913
## Specificity 0.99730 0.8937 0.9130
## Pos Pred Value 0.86054 0.9106 0.8883
## Neg Pred Value 0.99737 0.8906 0.9154
## Prevalence 0.01892 0.5441 0.4370
## Detection Rate 0.01633 0.4941 0.3895
## Detection Prevalence 0.01898 0.5425 0.4385
## Balanced Accuracy 0.93039 0.9009 0.9021
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13941, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9337628 0.8699625
## 2 extratrees 0.8936739 0.7890762
## 7 gini 0.9426073 0.8877627
## 7 extratrees 0.9452549 0.8927048
## 12 gini 0.9316314 0.8663912
## 12 extratrees 0.9504841 0.9030756
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 264 5 5
## 1 23 8141 446
## 2 6 282 6318
##
## Overall Statistics
##
## Accuracy : 0.9505
## 95% CI : (0.9469, 0.9538)
## No Information Rate : 0.5441
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9031
##
## Mcnemar's Test P-Value : 1.581e-10
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.90102 0.9659 0.9334
## Specificity 0.99934 0.9336 0.9670
## Pos Pred Value 0.96350 0.9455 0.9564
## Neg Pred Value 0.99809 0.9583 0.9492
## Prevalence 0.01892 0.5441 0.4370
## Detection Rate 0.01704 0.5256 0.4079
## Detection Prevalence 0.01769 0.5558 0.4265
## Balanced Accuracy 0.95018 0.9498 0.9502
## [1] "----------------------------------------------------------------------------"
## [1] "1 26 345"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13941, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9219493 0.8106317
## 2 0.9016800 0.7620835
## 3 0.9063923 0.7707197
## 4 0.8914798 0.7338248
## 5 0.8941913 0.7394421
## 6 0.8840563 0.7131525
## 7 0.8828942 0.7085669
## 8 0.8791504 0.6990369
## 9 0.8746950 0.6863672
## 10 0.8715965 0.6776932
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 257 22 14
## 1 24 3436 585
## 2 12 552 10588
##
## Overall Statistics
##
## Accuracy : 0.9219
## 95% CI : (0.9176, 0.9261)
## No Information Rate : 0.7222
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8106
##
## Mcnemar's Test P-Value : 0.7533
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87713 0.8569 0.9465
## Specificity 0.99763 0.9470 0.8689
## Pos Pred Value 0.87713 0.8494 0.9494
## Neg Pred Value 0.99763 0.9498 0.8619
## Prevalence 0.01892 0.2589 0.7222
## Detection Rate 0.01659 0.2218 0.6835
## Detection Prevalence 0.01892 0.2611 0.7199
## Balanced Accuracy 0.93738 0.9019 0.9077
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13940, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9369272 0.8398899
## 2 extratrees 0.8961266 0.7187766
## 7 gini 0.9556485 0.8904595
## 7 extratrees 0.9525497 0.8814501
## 12 gini 0.9513877 0.8800858
## 12 extratrees 0.9554552 0.8893174
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 7, splitrule = gini
## and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 263 4 8
## 1 17 3590 229
## 2 13 416 10950
##
## Overall Statistics
##
## Accuracy : 0.9556
## 95% CI : (0.9523, 0.9588)
## No Information Rate : 0.7222
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8905
##
## Mcnemar's Test P-Value : 1.074e-13
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89761 0.8953 0.9788
## Specificity 0.99921 0.9786 0.9003
## Pos Pred Value 0.95636 0.9359 0.9623
## Neg Pred Value 0.99803 0.9640 0.9423
## Prevalence 0.01892 0.2589 0.7222
## Detection Rate 0.01698 0.2318 0.7069
## Detection Prevalence 0.01775 0.2476 0.7346
## Balanced Accuracy 0.94841 0.9369 0.9396
## [1] "----------------------------------------------------------------------------"
## [1] "1 34 256"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13942, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9429959 0.8464118
## 2 0.9289224 0.8084225
## 3 0.9350561 0.8218466
## 4 0.9264057 0.7975319
## 5 0.9268571 0.7966024
## 6 0.9229191 0.7845332
## 7 0.9215628 0.7791141
## 8 0.9174308 0.7672274
## 9 0.9166563 0.7635024
## 10 0.9148484 0.7581250
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 256 9 22
## 1 11 2983 406
## 2 26 409 11368
##
## Overall Statistics
##
## Accuracy : 0.943
## 95% CI : (0.9392, 0.9466)
## No Information Rate : 0.7615
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8464
##
## Mcnemar's Test P-Value : 0.909
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87372 0.8771 0.9637
## Specificity 0.99796 0.9655 0.8822
## Pos Pred Value 0.89199 0.8774 0.9631
## Neg Pred Value 0.99757 0.9654 0.8839
## Prevalence 0.01892 0.2196 0.7615
## Detection Rate 0.01653 0.1926 0.7339
## Detection Prevalence 0.01853 0.2195 0.7620
## Balanced Accuracy 0.93584 0.9213 0.9230
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13939, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9515177 0.8617314
## 2 extratrees 0.9245314 0.7736837
## 7 gini 0.9660425 0.9058740
## 7 extratrees 0.9644293 0.9007322
## 12 gini 0.9647517 0.9028484
## 12 extratrees 0.9664948 0.9069023
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 266 9 4
## 1 4 3000 87
## 2 23 392 11705
##
## Overall Statistics
##
## Accuracy : 0.9665
## 95% CI : (0.9635, 0.9693)
## No Information Rate : 0.7615
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9069
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.90785 0.8821 0.9923
## Specificity 0.99914 0.9925 0.8877
## Pos Pred Value 0.95341 0.9706 0.9658
## Neg Pred Value 0.99822 0.9677 0.9730
## Prevalence 0.01892 0.2196 0.7615
## Detection Rate 0.01717 0.1937 0.7556
## Detection Prevalence 0.01801 0.1995 0.7824
## Balanced Accuracy 0.95350 0.9373 0.9400
## [1] "----------------------------------------------------------------------------"
## [1] "1 35 246"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13940, 13942, 13940, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9013560 0.8089995
## 2 0.8752106 0.7583754
## 3 0.8812793 0.7699205
## 4 0.8681749 0.7442454
## 5 0.8679169 0.7436301
## 6 0.8586197 0.7254291
## 7 0.8564261 0.7209036
## 8 0.8506161 0.7097175
## 9 0.8490665 0.7063807
## 10 0.8429330 0.6944924
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 256 9 24
## 1 9 7328 707
## 2 28 751 6378
##
## Overall Statistics
##
## Accuracy : 0.9014
## 95% CI : (0.8966, 0.906)
## No Information Rate : 0.5221
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.809
##
## Mcnemar's Test P-Value : 0.6514
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87372 0.9060 0.8972
## Specificity 0.99783 0.9033 0.9071
## Pos Pred Value 0.88581 0.9110 0.8912
## Neg Pred Value 0.99757 0.8979 0.9123
## Prevalence 0.01892 0.5221 0.4589
## Detection Rate 0.01653 0.4731 0.4117
## Detection Prevalence 0.01866 0.5193 0.4620
## Balanced Accuracy 0.93577 0.9047 0.9021
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13941, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9339580 0.8715559
## 2 extratrees 0.8949646 0.7943071
## 7 gini 0.9433829 0.8901187
## 7 extratrees 0.9465466 0.8961824
## 12 gini 0.9367332 0.8772602
## 12 extratrees 0.9499036 0.9027704
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 263 3 5
## 1 1 7778 431
## 2 29 307 6673
##
## Overall Statistics
##
## Accuracy : 0.9499
## 95% CI : (0.9464, 0.9533)
## No Information Rate : 0.5221
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9028
##
## Mcnemar's Test P-Value : 1.936e-08
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89761 0.9617 0.9387
## Specificity 0.99947 0.9416 0.9599
## Pos Pred Value 0.97048 0.9474 0.9521
## Neg Pred Value 0.99803 0.9574 0.9486
## Prevalence 0.01892 0.5221 0.4589
## Detection Rate 0.01698 0.5021 0.4308
## Detection Prevalence 0.01750 0.5300 0.4525
## Balanced Accuracy 0.94854 0.9517 0.9493
## [1] "----------------------------------------------------------------------------"
## [1] "1 36 245"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13941, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9236938 0.8047422
## 2 0.9021949 0.7498060
## 3 0.9100714 0.7675910
## 4 0.8981279 0.7360986
## 5 0.9017430 0.7438741
## 6 0.8927056 0.7196431
## 7 0.8920591 0.7159529
## 8 0.8857323 0.6980919
## 9 0.8849571 0.6943025
## 10 0.8799219 0.6798737
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 253 5 30
## 1 7 3131 573
## 2 33 534 10924
##
## Overall Statistics
##
## Accuracy : 0.9237
## 95% CI : (0.9194, 0.9278)
## No Information Rate : 0.7442
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8048
##
## Mcnemar's Test P-Value : 0.6041
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.86348 0.8531 0.9477
## Specificity 0.99770 0.9509 0.8569
## Pos Pred Value 0.87847 0.8437 0.9507
## Neg Pred Value 0.99737 0.9542 0.8492
## Prevalence 0.01892 0.2369 0.7442
## Detection Rate 0.01633 0.2021 0.7052
## Detection Prevalence 0.01859 0.2396 0.7418
## Balanced Accuracy 0.93059 0.9020 0.9023
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13941, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9383476 0.8324575
## 2 extratrees 0.8908975 0.6790487
## 7 gini 0.9553257 0.8826213
## 7 extratrees 0.9513235 0.8705650
## 12 gini 0.9479016 0.8631859
## 12 extratrees 0.9557775 0.8834827
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 261 5 5
## 1 2 3207 185
## 2 30 458 11337
##
## Overall Statistics
##
## Accuracy : 0.9558
## 95% CI : (0.9524, 0.959)
## No Information Rate : 0.7442
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8835
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89078 0.8738 0.9835
## Specificity 0.99934 0.9842 0.8769
## Pos Pred Value 0.96310 0.9449 0.9587
## Neg Pred Value 0.99790 0.9617 0.9482
## Prevalence 0.01892 0.2369 0.7442
## Detection Rate 0.01685 0.2070 0.7319
## Detection Prevalence 0.01750 0.2191 0.7634
## Balanced Accuracy 0.94506 0.9290 0.9302
## [1] "----------------------------------------------------------------------------"
## [1] "1 45 236"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13939, 13942, 13940, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9204655 0.8145958
## 2 0.8952226 0.7569058
## 3 0.9063264 0.7802235
## 4 0.8930920 0.7487417
## 5 0.8948344 0.7514034
## 6 0.8839257 0.7249919
## 7 0.8834081 0.7221769
## 8 0.8800520 0.7147899
## 9 0.8757914 0.7022419
## 10 0.8716584 0.6925121
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 258 6 31
## 1 6 10288 569
## 2 29 591 3712
##
## Overall Statistics
##
## Accuracy : 0.9205
## 95% CI : (0.9161, 0.9247)
## No Information Rate : 0.7027
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8146
##
## Mcnemar's Test P-Value : 0.9224
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.88055 0.9452 0.8609
## Specificity 0.99757 0.8751 0.9445
## Pos Pred Value 0.87458 0.9471 0.8569
## Neg Pred Value 0.99770 0.8710 0.9462
## Prevalence 0.01892 0.7027 0.2784
## Detection Rate 0.01666 0.6642 0.2396
## Detection Prevalence 0.01904 0.7013 0.2797
## Balanced Accuracy 0.93906 0.9101 0.9027
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13942, 13940, 13940, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9383470 0.8507407
## 2 extratrees 0.8949652 0.7311994
## 7 gini 0.9519693 0.8862236
## 7 extratrees 0.9508074 0.8824352
## 12 gini 0.9502918 0.8822233
## 12 extratrees 0.9536477 0.8895574
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 259 1 9
## 1 3 10693 483
## 2 31 191 3820
##
## Overall Statistics
##
## Accuracy : 0.9536
## 95% CI : (0.9502, 0.9569)
## No Information Rate : 0.7027
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8896
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.88396 0.9824 0.8859
## Specificity 0.99934 0.8945 0.9801
## Pos Pred Value 0.96283 0.9565 0.9451
## Neg Pred Value 0.99777 0.9555 0.9570
## Prevalence 0.01892 0.7027 0.2784
## Detection Rate 0.01672 0.6903 0.2466
## Detection Prevalence 0.01737 0.7217 0.2609
## Balanced Accuracy 0.94165 0.9384 0.9330
## [1] "----------------------------------------------------------------------------"
## [1] "1 46 235"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13940, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9005813 0.8041742
## 2 0.8723040 0.7486923
## 3 0.8821812 0.7674101
## 4 0.8687537 0.7408595
## 5 0.8672682 0.7374301
## 6 0.8544217 0.7118903
## 7 0.8550025 0.7124832
## 8 0.8458998 0.6943180
## 9 0.8437703 0.6898031
## 10 0.8377014 0.6773312
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 256 5 33
## 1 4 5729 732
## 2 33 733 7965
##
## Overall Statistics
##
## Accuracy : 0.9006
## 95% CI : (0.8958, 0.9052)
## No Information Rate : 0.5636
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8042
##
## Mcnemar's Test P-Value : 0.9904
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87372 0.8859 0.9124
## Specificity 0.99750 0.9184 0.8867
## Pos Pred Value 0.87075 0.8862 0.9123
## Neg Pred Value 0.99757 0.9182 0.8868
## Prevalence 0.01892 0.4175 0.5636
## Detection Rate 0.01653 0.3699 0.5142
## Detection Prevalence 0.01898 0.4174 0.5637
## Balanced Accuracy 0.93561 0.9022 0.8995
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9335690 0.8679184
## 2 extratrees 0.8882482 0.7738845
## 7 gini 0.9418979 0.8850729
## 7 extratrees 0.9446097 0.8900750
## 12 gini 0.9342155 0.8700103
## 12 extratrees 0.9475794 0.8961085
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 246 0 5
## 1 5 5998 291
## 2 42 469 8434
##
## Overall Statistics
##
## Accuracy : 0.9476
## 95% CI : (0.944, 0.951)
## No Information Rate : 0.5636
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8961
##
## Mcnemar's Test P-Value : 2.421e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.83959 0.9275 0.9661
## Specificity 0.99967 0.9672 0.9244
## Pos Pred Value 0.98008 0.9530 0.9429
## Neg Pred Value 0.99692 0.9490 0.9548
## Prevalence 0.01892 0.4175 0.5636
## Detection Rate 0.01588 0.3872 0.5445
## Detection Prevalence 0.01620 0.4063 0.5775
## Balanced Accuracy 0.91963 0.9473 0.9453
## [1] "----------------------------------------------------------------------------"
## [1] "1 56 234"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13939, 13942, 13939, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9452540 0.8676650
## 2 0.9322123 0.8360238
## 3 0.9366667 0.8449950
## 4 0.9311783 0.8310795
## 5 0.9320174 0.8319397
## 6 0.9270473 0.8187843
## 7 0.9268527 0.8179131
## 8 0.9222060 0.8058301
## 9 0.9222050 0.8048997
## 10 0.9191053 0.7971084
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 258 2 38
## 1 2 10756 377
## 2 33 396 3628
##
## Overall Statistics
##
## Accuracy : 0.9453
## 95% CI : (0.9416, 0.9488)
## No Information Rate : 0.7201
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8677
##
## Mcnemar's Test P-Value : 0.8449
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.88055 0.9643 0.8974
## Specificity 0.99737 0.9126 0.9625
## Pos Pred Value 0.86577 0.9660 0.8943
## Neg Pred Value 0.99770 0.9086 0.9637
## Prevalence 0.01892 0.7201 0.2610
## Detection Rate 0.01666 0.6944 0.2342
## Detection Prevalence 0.01924 0.7189 0.2619
## Balanced Accuracy 0.93896 0.9385 0.9299
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13940, 13942, 13941, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9553252 0.8880077
## 2 extratrees 0.9320842 0.8248334
## 7 gini 0.9681717 0.9215827
## 7 extratrees 0.9651375 0.9135792
## 12 gini 0.9646856 0.9132197
## 12 extratrees 0.9686884 0.9226879
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 256 1 7
## 1 2 11057 344
## 2 35 96 3692
##
## Overall Statistics
##
## Accuracy : 0.9687
## 95% CI : (0.9658, 0.9714)
## No Information Rate : 0.7201
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9227
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87372 0.9913 0.9132
## Specificity 0.99947 0.9202 0.9886
## Pos Pred Value 0.96970 0.9697 0.9657
## Neg Pred Value 0.99757 0.9763 0.9699
## Prevalence 0.01892 0.7201 0.2610
## Detection Rate 0.01653 0.7138 0.2383
## Detection Prevalence 0.01704 0.7362 0.2468
## Balanced Accuracy 0.93660 0.9558 0.9509
## [1] "----------------------------------------------------------------------------"
## [1] "2 13 456"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13940, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9856039 0.9054955
## 2 0.9795353 0.8651191
## 3 0.9814719 0.8767209
## 4 0.9785023 0.8561079
## 5 0.9780504 0.8521696
## 6 0.9761138 0.8383263
## 7 0.9756612 0.8344425
## 8 0.9749512 0.8286273
## 9 0.9746932 0.8260645
## 10 0.9730794 0.8141102
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 572 32 53
## 1 31 539 44
## 2 39 24 14156
##
## Overall Statistics
##
## Accuracy : 0.9856
## 95% CI : (0.9836, 0.9874)
## No Information Rate : 0.9201
## P-Value [Acc > NIR] : < 2e-16
##
## Kappa : 0.9053
##
## Mcnemar's Test P-Value : 0.04542
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89097 0.90588 0.9932
## Specificity 0.99428 0.99496 0.9491
## Pos Pred Value 0.87062 0.87785 0.9956
## Neg Pred Value 0.99528 0.99624 0.9237
## Prevalence 0.04145 0.03841 0.9201
## Detection Rate 0.03693 0.03480 0.9139
## Detection Prevalence 0.04241 0.03964 0.9179
## Balanced Accuracy 0.94262 0.95042 0.9711
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13942, 13941, 13942, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9859270 0.8999613
## 2 extratrees 0.9788899 0.8419836
## 7 gini 0.9908976 0.9375618
## 7 extratrees 0.9916074 0.9425974
## 12 gini 0.9899935 0.9319246
## 12 extratrees 0.9916721 0.9432019
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 578 19 9
## 1 16 548 9
## 2 48 28 14235
##
## Overall Statistics
##
## Accuracy : 0.9917
## 95% CI : (0.9901, 0.993)
## No Information Rate : 0.9201
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9433
##
## Mcnemar's Test P-Value : 5.33e-08
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.90031 0.92101 0.9987
## Specificity 0.99811 0.99832 0.9386
## Pos Pred Value 0.95380 0.95637 0.9947
## Neg Pred Value 0.99570 0.99685 0.9847
## Prevalence 0.04145 0.03841 0.9201
## Detection Rate 0.03731 0.03538 0.9190
## Detection Prevalence 0.03912 0.03699 0.9239
## Balanced Accuracy 0.94921 0.95966 0.9686
## [1] "----------------------------------------------------------------------------"
## [1] "2 14 356"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9389295 0.8492151
## 2 0.9252426 0.8150750
## 3 0.9315052 0.8278995
## 4 0.9232417 0.8066755
## 5 0.9229823 0.8047304
## 6 0.9180760 0.7914785
## 7 0.9174294 0.7882930
## 8 0.9140728 0.7788794
## 9 0.9122010 0.7731417
## 10 0.9076168 0.7603168
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 572 73 31
## 1 52 2943 396
## 2 18 376 11029
##
## Overall Statistics
##
## Accuracy : 0.9389
## 95% CI : (0.935, 0.9426)
## No Information Rate : 0.7396
## P-Value [Acc > NIR] : < 2e-16
##
## Kappa : 0.8492
##
## Mcnemar's Test P-Value : 0.05768
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89097 0.8676 0.9627
## Specificity 0.99300 0.9630 0.9023
## Pos Pred Value 0.84615 0.8679 0.9655
## Neg Pred Value 0.99527 0.9629 0.8950
## Prevalence 0.04145 0.2190 0.7396
## Detection Rate 0.03693 0.1900 0.7120
## Detection Prevalence 0.04364 0.2189 0.7374
## Balanced Accuracy 0.94198 0.9153 0.9325
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9508720 0.8723246
## 2 extratrees 0.9266634 0.8012042
## 7 gini 0.9637829 0.9080534
## 7 extratrees 0.9646225 0.9095826
## 12 gini 0.9601674 0.8991730
## 12 extratrees 0.9672695 0.9167384
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 604 16 6
## 1 31 3005 76
## 2 7 371 11374
##
## Overall Statistics
##
## Accuracy : 0.9673
## 95% CI : (0.9643, 0.97)
## No Information Rate : 0.7396
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9168
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.94081 0.8859 0.9928
## Specificity 0.99852 0.9912 0.9063
## Pos Pred Value 0.96486 0.9656 0.9678
## Neg Pred Value 0.99744 0.9687 0.9781
## Prevalence 0.04145 0.2190 0.7396
## Detection Rate 0.03899 0.1940 0.7343
## Detection Prevalence 0.04041 0.2009 0.7587
## Balanced Accuracy 0.96966 0.9385 0.9496
## [1] "----------------------------------------------------------------------------"
## [1] "2 15 346"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8949642 0.8041470
## 2 0.8730783 0.7634114
## 3 0.8770172 0.7706244
## 4 0.8643641 0.7468485
## 5 0.8625564 0.7434032
## 6 0.8536483 0.7269599
## 7 0.8497101 0.7193596
## 8 0.8430603 0.7069301
## 9 0.8404144 0.7018235
## 10 0.8357656 0.6931187
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 572 42 54
## 1 33 7280 704
## 2 37 757 6011
##
## Overall Statistics
##
## Accuracy : 0.895
## 95% CI : (0.89, 0.8998)
## No Information Rate : 0.5216
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8041
##
## Mcnemar's Test P-Value : 0.1032
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89097 0.9011 0.8880
## Specificity 0.99353 0.9006 0.9090
## Pos Pred Value 0.85629 0.9081 0.8833
## Neg Pred Value 0.99528 0.8931 0.9127
## Prevalence 0.04145 0.5216 0.4370
## Detection Rate 0.03693 0.4700 0.3881
## Detection Prevalence 0.04312 0.5176 0.4393
## Balanced Accuracy 0.94225 0.9008 0.8985
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13940, 13942, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9335054 0.8752443
## 2 extratrees 0.8927033 0.7965489
## 7 gini 0.9449325 0.8969425
## 7 extratrees 0.9464816 0.8996961
## 12 gini 0.9366698 0.8815609
## 12 extratrees 0.9498386 0.9060689
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 605 16 6
## 1 12 7785 440
## 2 25 278 6323
##
## Overall Statistics
##
## Accuracy : 0.9498
## 95% CI : (0.9463, 0.9532)
## No Information Rate : 0.5216
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9061
##
## Mcnemar's Test P-Value : 1.461e-10
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.94237 0.9636 0.9341
## Specificity 0.99852 0.9390 0.9653
## Pos Pred Value 0.96491 0.9451 0.9543
## Neg Pred Value 0.99751 0.9595 0.9497
## Prevalence 0.04145 0.5216 0.4370
## Detection Rate 0.03906 0.5026 0.4082
## Detection Prevalence 0.04048 0.5318 0.4278
## Balanced Accuracy 0.97044 0.9513 0.9497
## [1] "----------------------------------------------------------------------------"
## [1] "2 16 345"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13939, 13940, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9207860 0.8126305
## 2 0.8996758 0.7626148
## 3 0.9073596 0.7783837
## 4 0.8956739 0.7503523
## 5 0.8956095 0.7485429
## 6 0.8868944 0.7266064
## 7 0.8868293 0.7252278
## 8 0.8819869 0.7135613
## 9 0.8770799 0.6996740
## 10 0.8730135 0.6886997
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 574 27 63
## 1 22 3116 551
## 2 46 518 10573
##
## Overall Statistics
##
## Accuracy : 0.9208
## 95% CI : (0.9164, 0.925)
## No Information Rate : 0.7222
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8126
##
## Mcnemar's Test P-Value : 0.2426
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89408 0.8511 0.9451
## Specificity 0.99394 0.9516 0.8689
## Pos Pred Value 0.86446 0.8447 0.9494
## Neg Pred Value 0.99541 0.9538 0.8589
## Prevalence 0.04145 0.2363 0.7222
## Detection Rate 0.03706 0.2012 0.6826
## Detection Prevalence 0.04287 0.2382 0.7190
## Balanced Accuracy 0.94401 0.9013 0.9070
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13940, 13941, 13940, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9364112 0.8412282
## 2 extratrees 0.8950933 0.7200445
## 7 gini 0.9551317 0.8916552
## 7 extratrees 0.9517105 0.8820843
## 12 gini 0.9505485 0.8807918
## 12 extratrees 0.9546801 0.8898852
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 7, splitrule = gini
## and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 585 17 18
## 1 6 3259 218
## 2 51 385 10951
##
## Overall Statistics
##
## Accuracy : 0.9551
## 95% CI : (0.9518, 0.9583)
## No Information Rate : 0.7222
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8917
##
## Mcnemar's Test P-Value : 1.62e-14
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.91121 0.8902 0.9789
## Specificity 0.99764 0.9811 0.8987
## Pos Pred Value 0.94355 0.9357 0.9617
## Neg Pred Value 0.99617 0.9665 0.9425
## Prevalence 0.04145 0.2363 0.7222
## Detection Rate 0.03777 0.2104 0.7070
## Detection Prevalence 0.04003 0.2249 0.7351
## Balanced Accuracy 0.95443 0.9356 0.9388
## [1] "----------------------------------------------------------------------------"
## [1] "2 34 156"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13939, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9415084 0.8557936
## 2 0.9282099 0.8228002
## 3 0.9320193 0.8296883
## 4 0.9245301 0.8102340
## 5 0.9233682 0.8063508
## 6 0.9198173 0.7967822
## 7 0.9182680 0.7921369
## 8 0.9137487 0.7801793
## 9 0.9152975 0.7833914
## 10 0.9116174 0.7730748
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 571 39 49
## 1 38 3001 386
## 2 33 361 11012
##
## Overall Statistics
##
## Accuracy : 0.9415
## 95% CI : (0.9377, 0.9452)
## No Information Rate : 0.739
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8558
##
## Mcnemar's Test P-Value : 0.2645
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.88941 0.8824 0.9620
## Specificity 0.99407 0.9649 0.9025
## Pos Pred Value 0.86646 0.8762 0.9655
## Neg Pred Value 0.99521 0.9668 0.8935
## Prevalence 0.04145 0.2196 0.7390
## Detection Rate 0.03686 0.1937 0.7109
## Detection Prevalence 0.04254 0.2211 0.7363
## Balanced Accuracy 0.94174 0.9237 0.9323
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13940, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9515175 0.8743674
## 2 extratrees 0.9249198 0.7964214
## 7 gini 0.9655910 0.9128665
## 7 extratrees 0.9638480 0.9078828
## 12 gini 0.9632022 0.9071993
## 12 extratrees 0.9668824 0.9159050
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 601 8 16
## 1 27 3023 78
## 2 14 370 11353
##
## Overall Statistics
##
## Accuracy : 0.9669
## 95% CI : (0.9639, 0.9696)
## No Information Rate : 0.739
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.916
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.93614 0.8889 0.9918
## Specificity 0.99838 0.9913 0.9050
## Pos Pred Value 0.96160 0.9664 0.9673
## Neg Pred Value 0.99724 0.9694 0.9750
## Prevalence 0.04145 0.2196 0.7390
## Detection Rate 0.03880 0.1952 0.7329
## Detection Prevalence 0.04035 0.2019 0.7577
## Balanced Accuracy 0.96726 0.9401 0.9484
## [1] "----------------------------------------------------------------------------"
## [1] "2 35 146"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8985803 0.8109151
## 2 0.8736610 0.7643347
## 3 0.8778568 0.7720032
## 4 0.8653332 0.7487951
## 5 0.8632030 0.7445290
## 6 0.8523571 0.7244508
## 7 0.8489361 0.7176403
## 8 0.8417700 0.7043785
## 9 0.8406076 0.7020345
## 10 0.8366691 0.6944986
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 582 25 66
## 1 15 7323 680
## 2 45 740 6014
##
## Overall Statistics
##
## Accuracy : 0.8986
## 95% CI : (0.8937, 0.9033)
## No Information Rate : 0.5221
## P-Value [Acc > NIR] : < 2e-16
##
## Kappa : 0.8109
##
## Mcnemar's Test P-Value : 0.02918
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.90654 0.9054 0.8896
## Specificity 0.99387 0.9061 0.9101
## Pos Pred Value 0.86478 0.9133 0.8845
## Neg Pred Value 0.99595 0.8976 0.9142
## Prevalence 0.04145 0.5221 0.4364
## Detection Rate 0.03757 0.4728 0.3883
## Detection Prevalence 0.04345 0.5176 0.4389
## Balanced Accuracy 0.95021 0.9058 0.8999
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13942, 13942, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9322785 0.8727918
## 2 extratrees 0.8927704 0.7965550
## 7 gini 0.9431893 0.8937367
## 7 extratrees 0.9472562 0.9011295
## 12 gini 0.9347319 0.8779103
## 12 extratrees 0.9501612 0.9066199
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 599 6 15
## 1 8 7800 426
## 2 35 282 6319
##
## Overall Statistics
##
## Accuracy : 0.9502
## 95% CI : (0.9466, 0.9535)
## No Information Rate : 0.5221
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9066
##
## Mcnemar's Test P-Value : 3.479e-08
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.93302 0.9644 0.9348
## Specificity 0.99859 0.9414 0.9637
## Pos Pred Value 0.96613 0.9473 0.9522
## Neg Pred Value 0.99711 0.9603 0.9502
## Prevalence 0.04145 0.5221 0.4364
## Detection Rate 0.03867 0.5036 0.4079
## Detection Prevalence 0.04003 0.5316 0.4284
## Balanced Accuracy 0.96580 0.9529 0.9492
## [1] "----------------------------------------------------------------------------"
## [1] "2 36 145"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13942, 13940, 13940, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9178833 0.8061207
## 2 0.8971600 0.7577858
## 3 0.9026475 0.7681082
## 4 0.8900594 0.7374250
## 5 0.8908985 0.7383578
## 6 0.8841850 0.7226276
## 7 0.8808909 0.7126707
## 8 0.8758551 0.6997940
## 9 0.8724979 0.6902647
## 10 0.8688184 0.6808357
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 570 11 78
## 1 7 3129 581
## 2 65 530 10519
##
## Overall Statistics
##
## Accuracy : 0.9179
## 95% CI : (0.9134, 0.9222)
## No Information Rate : 0.7216
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8062
##
## Mcnemar's Test P-Value : 0.2203
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.88785 0.8526 0.9410
## Specificity 0.99401 0.9503 0.8620
## Pos Pred Value 0.86495 0.8418 0.9465
## Neg Pred Value 0.99515 0.9540 0.8494
## Prevalence 0.04145 0.2369 0.7216
## Detection Rate 0.03680 0.2020 0.6791
## Detection Prevalence 0.04254 0.2400 0.7175
## Balanced Accuracy 0.94093 0.9014 0.9015
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13941, 13942, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9346040 0.8363756
## 2 extratrees 0.8859916 0.6924613
## 7 gini 0.9504849 0.8797338
## 7 extratrees 0.9491290 0.8753778
## 12 gini 0.9467402 0.8706837
## 12 extratrees 0.9524219 0.8841719
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 589 2 18
## 1 3 3200 196
## 2 50 468 10964
##
## Overall Statistics
##
## Accuracy : 0.9524
## 95% CI : (0.949, 0.9557)
## No Information Rate : 0.7216
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8844
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.91745 0.8719 0.9809
## Specificity 0.99865 0.9832 0.8799
## Pos Pred Value 0.96716 0.9415 0.9549
## Neg Pred Value 0.99644 0.9611 0.9466
## Prevalence 0.04145 0.2369 0.7216
## Detection Rate 0.03802 0.2066 0.7078
## Detection Prevalence 0.03932 0.2194 0.7413
## Balanced Accuracy 0.95805 0.9275 0.9304
## [1] "----------------------------------------------------------------------------"
## [1] "2 45 136"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13942, 13940, 13941, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9177531 0.8132118
## 2 0.8980634 0.7686947
## 3 0.9035512 0.7787943
## 4 0.8885729 0.7447650
## 5 0.8933494 0.7543626
## 6 0.8820515 0.7283425
## 7 0.8790185 0.7185431
## 8 0.8752093 0.7092709
## 9 0.8724331 0.7017976
## 10 0.8688834 0.6932710
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 571 52 37
## 1 39 10267 548
## 2 32 566 3378
##
## Overall Statistics
##
## Accuracy : 0.9178
## 95% CI : (0.9133, 0.922)
## No Information Rate : 0.7027
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8131
##
## Mcnemar's Test P-Value : 0.4734
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.88941 0.9432 0.8524
## Specificity 0.99401 0.8725 0.9481
## Pos Pred Value 0.86515 0.9459 0.8496
## Neg Pred Value 0.99521 0.8667 0.9492
## Prevalence 0.04145 0.7027 0.2558
## Detection Rate 0.03686 0.6628 0.2181
## Detection Prevalence 0.04261 0.7007 0.2567
## Balanced Accuracy 0.94171 0.9079 0.9003
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13941, 13941, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9366689 0.8493188
## 2 extratrees 0.8915420 0.7258372
## 7 gini 0.9531956 0.8915727
## 7 extratrees 0.9506775 0.8847482
## 12 gini 0.9510651 0.8867104
## 12 extratrees 0.9530664 0.8907996
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 7, splitrule = gini
## and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 593 13 15
## 1 39 10656 432
## 2 10 216 3516
##
## Overall Statistics
##
## Accuracy : 0.9532
## 95% CI : (0.9498, 0.9565)
## No Information Rate : 0.7027
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8916
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.92368 0.9790 0.8872
## Specificity 0.99811 0.8977 0.9804
## Pos Pred Value 0.95491 0.9577 0.9396
## Neg Pred Value 0.99670 0.9475 0.9620
## Prevalence 0.04145 0.7027 0.2558
## Detection Rate 0.03828 0.6879 0.2270
## Detection Prevalence 0.04009 0.7183 0.2416
## Balanced Accuracy 0.96090 0.9383 0.9338
## [1] "----------------------------------------------------------------------------"
## [1] "2 46 135"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13940, 13941, 13941, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8992918 0.8108597
## 2 0.8741132 0.7635785
## 3 0.8778583 0.7703283
## 4 0.8649461 0.7461470
## 5 0.8633342 0.7427242
## 6 0.8544902 0.7261981
## 7 0.8495194 0.7165997
## 8 0.8432568 0.7047483
## 9 0.8398362 0.6982071
## 10 0.8355752 0.6895983
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 578 39 59
## 1 27 5739 709
## 2 37 689 7613
##
## Overall Statistics
##
## Accuracy : 0.8993
## 95% CI : (0.8944, 0.904)
## No Information Rate : 0.5411
## P-Value [Acc > NIR] : < 2e-16
##
## Kappa : 0.8108
##
## Mcnemar's Test P-Value : 0.05731
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.90031 0.8874 0.9084
## Specificity 0.99340 0.9184 0.8979
## Pos Pred Value 0.85503 0.8863 0.9129
## Neg Pred Value 0.99568 0.9192 0.8926
## Prevalence 0.04145 0.4175 0.5411
## Detection Rate 0.03731 0.3705 0.4915
## Detection Prevalence 0.04364 0.4180 0.5383
## Balanced Accuracy 0.94686 0.9029 0.9031
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13940, 13941, 13940, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9327955 0.8725492
## 2 extratrees 0.8880526 0.7842797
## 7 gini 0.9430602 0.8925526
## 7 extratrees 0.9448676 0.8956576
## 12 gini 0.9374433 0.8819784
## 12 extratrees 0.9474504 0.9006777
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 598 6 11
## 1 21 6003 295
## 2 23 458 8075
##
## Overall Statistics
##
## Accuracy : 0.9474
## 95% CI : (0.9438, 0.9509)
## No Information Rate : 0.5411
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9007
##
## Mcnemar's Test P-Value : 2.289e-10
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.93146 0.9283 0.9635
## Specificity 0.99886 0.9650 0.9323
## Pos Pred Value 0.97236 0.9500 0.9438
## Neg Pred Value 0.99704 0.9494 0.9559
## Prevalence 0.04145 0.4175 0.5411
## Detection Rate 0.03861 0.3875 0.5213
## Detection Prevalence 0.03970 0.4079 0.5524
## Balanced Accuracy 0.96516 0.9466 0.9479
## [1] "----------------------------------------------------------------------------"
## [1] "2 56 134"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13942, 13939, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9428022 0.8650640
## 2 0.9285348 0.8312834
## 3 0.9321493 0.8377285
## 4 0.9245316 0.8191893
## 5 0.9254347 0.8199656
## 6 0.9209160 0.8082910
## 7 0.9198846 0.8053442
## 8 0.9164622 0.7963059
## 9 0.9171716 0.7971990
## 10 0.9122649 0.7850007
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 577 19 73
## 1 11 10764 358
## 2 54 371 3263
##
## Overall Statistics
##
## Accuracy : 0.9428
## 95% CI : (0.939, 0.9464)
## No Information Rate : 0.7201
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8651
##
## Mcnemar's Test P-Value : 0.1572
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89875 0.9650 0.8833
## Specificity 0.99380 0.9149 0.9640
## Pos Pred Value 0.86248 0.9669 0.8848
## Neg Pred Value 0.99561 0.9105 0.9635
## Prevalence 0.04145 0.7201 0.2385
## Detection Rate 0.03725 0.6949 0.2107
## Detection Prevalence 0.04319 0.7187 0.2381
## Balanced Accuracy 0.94628 0.9400 0.9236
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13940, 13940, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9535191 0.8859737
## 2 extratrees 0.9264045 0.8126124
## 7 gini 0.9668820 0.9203505
## 7 extratrees 0.9654622 0.9164190
## 12 gini 0.9644940 0.9147834
## 12 extratrees 0.9681737 0.9232827
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 602 6 16
## 1 3 11060 343
## 2 37 88 3335
##
## Overall Statistics
##
## Accuracy : 0.9682
## 95% CI : (0.9653, 0.9709)
## No Information Rate : 0.7201
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9233
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.93769 0.9916 0.9028
## Specificity 0.99852 0.9202 0.9894
## Pos Pred Value 0.96474 0.9697 0.9639
## Neg Pred Value 0.99731 0.9770 0.9702
## Prevalence 0.04145 0.7201 0.2385
## Detection Rate 0.03886 0.7140 0.2153
## Detection Prevalence 0.04028 0.7363 0.2234
## Balanced Accuracy 0.96811 0.9559 0.9461
## [1] "----------------------------------------------------------------------------"
## [1] "3 12 456"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13942, 13940, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9874758 0.9169031
## 2 0.9839895 0.8932119
## 3 0.9855387 0.9032746
## 4 0.9819233 0.8786485
## 5 0.9830853 0.8855905
## 6 0.9805027 0.8678730
## 7 0.9796636 0.8611427
## 8 0.9781149 0.8506923
## 9 0.9774690 0.8445550
## 10 0.9770170 0.8404403
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 265 12 31
## 1 14 873 64
## 2 23 50 14158
##
## Overall Statistics
##
## Accuracy : 0.9875
## 95% CI : (0.9856, 0.9892)
## No Information Rate : 0.9201
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.9168
##
## Mcnemar's Test P-Value : 0.3827
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87748 0.93369 0.9933
## Specificity 0.99717 0.99464 0.9410
## Pos Pred Value 0.86039 0.91798 0.9949
## Neg Pred Value 0.99756 0.99574 0.9245
## Prevalence 0.01950 0.06036 0.9201
## Detection Rate 0.01711 0.05636 0.9140
## Detection Prevalence 0.01988 0.06139 0.9187
## Balanced Accuracy 0.93733 0.96417 0.9672
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13942, 13941, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9869589 0.9068708
## 2 extratrees 0.9784370 0.8376276
## 7 gini 0.9908323 0.9366460
## 7 extratrees 0.9925105 0.9485103
## 12 gini 0.9901869 0.9327936
## 12 extratrees 0.9926398 0.9496670
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 258 5 3
## 1 17 886 18
## 2 27 44 14232
##
## Overall Statistics
##
## Accuracy : 0.9926
## 95% CI : (0.9912, 0.9939)
## No Information Rate : 0.9201
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9497
##
## Mcnemar's Test P-Value : 5.46e-08
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.85430 0.94759 0.9985
## Specificity 0.99947 0.99760 0.9426
## Pos Pred Value 0.96992 0.96200 0.9950
## Neg Pred Value 0.99711 0.99664 0.9823
## Prevalence 0.01950 0.06036 0.9201
## Detection Rate 0.01666 0.05720 0.9188
## Detection Prevalence 0.01717 0.05946 0.9234
## Balanced Accuracy 0.92689 0.97259 0.9706
## [1] "----------------------------------------------------------------------------"
## [1] "3 14 256"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13941, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9420268 0.8440218
## 2 0.9255648 0.7994158
## 3 0.9319549 0.8133516
## 4 0.9251767 0.7937615
## 5 0.9257577 0.7939737
## 6 0.9202701 0.7779535
## 7 0.9198826 0.7753614
## 8 0.9178158 0.7688370
## 9 0.9152342 0.7596282
## 10 0.9133622 0.7532809
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 272 29 20
## 1 19 2945 401
## 2 11 418 11375
##
## Overall Statistics
##
## Accuracy : 0.942
## 95% CI : (0.9382, 0.9457)
## No Information Rate : 0.7615
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.844
##
## Mcnemar's Test P-Value : 0.1682
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.90066 0.8682 0.9643
## Specificity 0.99677 0.9653 0.8839
## Pos Pred Value 0.84735 0.8752 0.9637
## Neg Pred Value 0.99802 0.9631 0.8858
## Prevalence 0.01950 0.2190 0.7615
## Detection Rate 0.01756 0.1901 0.7343
## Detection Prevalence 0.02072 0.2172 0.7620
## Balanced Accuracy 0.94872 0.9168 0.9241
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9504851 0.8585932
## 2 extratrees 0.9238873 0.7716491
## 7 gini 0.9640418 0.9000715
## 7 extratrees 0.9634611 0.8978881
## 12 gini 0.9603617 0.8900222
## 12 extratrees 0.9663664 0.9064713
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 266 3 1
## 1 21 2996 88
## 2 15 393 11707
##
## Overall Statistics
##
## Accuracy : 0.9664
## 95% CI : (0.9634, 0.9691)
## No Information Rate : 0.7615
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9066
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.88079 0.8833 0.9925
## Specificity 0.99974 0.9910 0.8896
## Pos Pred Value 0.98519 0.9649 0.9663
## Neg Pred Value 0.99763 0.9680 0.9736
## Prevalence 0.01950 0.2190 0.7615
## Detection Rate 0.01717 0.1934 0.7558
## Detection Prevalence 0.01743 0.2005 0.7821
## Balanced Accuracy 0.94027 0.9371 0.9410
## [1] "----------------------------------------------------------------------------"
## [1] "3 15 246"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13942, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8993535 0.8054045
## 2 0.8714636 0.7514294
## 3 0.8788876 0.7657736
## 4 0.8651382 0.7392037
## 5 0.8683644 0.7455511
## 6 0.8609415 0.7310483
## 7 0.8566812 0.7227039
## 8 0.8486758 0.7070864
## 9 0.8467391 0.7033896
## 10 0.8418981 0.6938403
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 265 18 24
## 1 14 7311 730
## 2 23 750 6355
##
## Overall Statistics
##
## Accuracy : 0.8994
## 95% CI : (0.8945, 0.904)
## No Information Rate : 0.5216
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8054
##
## Mcnemar's Test P-Value : 0.8515
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87748 0.9049 0.8939
## Specificity 0.99723 0.8996 0.9078
## Pos Pred Value 0.86319 0.9076 0.8916
## Neg Pred Value 0.99756 0.8967 0.9098
## Prevalence 0.01950 0.5216 0.4589
## Detection Rate 0.01711 0.4720 0.4103
## Detection Prevalence 0.01982 0.5200 0.4602
## Balanced Accuracy 0.93736 0.9023 0.9009
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13942, 13941, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9327312 0.8692779
## 2 extratrees 0.8929629 0.7908406
## 7 gini 0.9424797 0.8883148
## 7 extratrees 0.9469983 0.8971712
## 12 gini 0.9382187 0.8800577
## 12 extratrees 0.9504847 0.9039716
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 267 2 4
## 1 12 7771 420
## 2 23 306 6685
##
## Overall Statistics
##
## Accuracy : 0.9505
## 95% CI : (0.9469, 0.9538)
## No Information Rate : 0.5216
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.904
##
## Mcnemar's Test P-Value : 2.31e-08
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.88411 0.9619 0.9404
## Specificity 0.99960 0.9417 0.9607
## Pos Pred Value 0.97802 0.9473 0.9531
## Neg Pred Value 0.99770 0.9577 0.9500
## Prevalence 0.01950 0.5216 0.4589
## Detection Rate 0.01724 0.5017 0.4316
## Detection Prevalence 0.01762 0.5296 0.4528
## Balanced Accuracy 0.94186 0.9518 0.9506
## [1] "----------------------------------------------------------------------------"
## [1] "3 16 245"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13941, 13940, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9241428 0.8061735
## 2 0.9062619 0.7612224
## 3 0.9105860 0.7689827
## 4 0.8975464 0.7344976
## 5 0.9014198 0.7433450
## 6 0.8925758 0.7201536
## 7 0.8917364 0.7155327
## 8 0.8877342 0.7046009
## 9 0.8856678 0.6974127
## 10 0.8803091 0.6830634
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 268 6 41
## 1 6 3120 559
## 2 28 535 10927
##
## Overall Statistics
##
## Accuracy : 0.9241
## 95% CI : (0.9199, 0.9283)
## No Information Rate : 0.7442
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8062
##
## Mcnemar's Test P-Value : 0.3954
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.88742 0.8522 0.9479
## Specificity 0.99691 0.9522 0.8579
## Pos Pred Value 0.85079 0.8467 0.9510
## Neg Pred Value 0.99776 0.9542 0.8500
## Prevalence 0.01950 0.2363 0.7442
## Detection Rate 0.01730 0.2014 0.7054
## Detection Prevalence 0.02034 0.2379 0.7418
## Balanced Accuracy 0.94216 0.9022 0.9029
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13941, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9353133 0.8239951
## 2 extratrees 0.8910910 0.6798730
## 7 gini 0.9543585 0.8802316
## 7 extratrees 0.9506136 0.8688969
## 12 gini 0.9495812 0.8675928
## 12 extratrees 0.9555203 0.8827027
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 255 1 3
## 1 4 3207 185
## 2 43 453 11339
##
## Overall Statistics
##
## Accuracy : 0.9555
## 95% CI : (0.9522, 0.9587)
## No Information Rate : 0.7442
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8827
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.84437 0.8760 0.9837
## Specificity 0.99974 0.9840 0.8748
## Pos Pred Value 0.98456 0.9443 0.9581
## Neg Pred Value 0.99691 0.9625 0.9486
## Prevalence 0.01950 0.2363 0.7442
## Detection Rate 0.01646 0.2070 0.7320
## Detection Prevalence 0.01672 0.2192 0.7640
## Balanced Accuracy 0.92205 0.9300 0.9293
## [1] "----------------------------------------------------------------------------"
## [1] "3 24 156"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13940, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9444167 0.8597840
## 2 0.9296972 0.8223006
## 3 0.9324733 0.8274564
## 4 0.9260825 0.8102922
## 5 0.9277615 0.8134811
## 6 0.9231130 0.8011232
## 7 0.9242756 0.8034572
## 8 0.9192395 0.7898016
## 9 0.9194330 0.7890571
## 10 0.9163339 0.7805569
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 269 29 18
## 1 23 3337 406
## 2 10 375 11023
##
## Overall Statistics
##
## Accuracy : 0.9444
## 95% CI : (0.9407, 0.948)
## No Information Rate : 0.739
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8599
##
## Mcnemar's Test P-Value : 0.2398
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89073 0.8920 0.9630
## Specificity 0.99691 0.9635 0.9048
## Pos Pred Value 0.85127 0.8861 0.9663
## Neg Pred Value 0.99783 0.9655 0.8961
## Prevalence 0.01950 0.2415 0.7390
## Detection Rate 0.01737 0.2154 0.7116
## Detection Prevalence 0.02040 0.2431 0.7365
## Balanced Accuracy 0.94382 0.9277 0.9339
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13940, 13941, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9530677 0.8758028
## 2 extratrees 0.9305365 0.8101922
## 7 gini 0.9664951 0.9133285
## 7 extratrees 0.9655269 0.9101746
## 12 gini 0.9598455 0.8961432
## 12 extratrees 0.9673340 0.9152470
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 263 4 2
## 1 21 3373 97
## 2 18 364 11348
##
## Overall Statistics
##
## Accuracy : 0.9673
## 95% CI : (0.9644, 0.9701)
## No Information Rate : 0.739
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9153
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87086 0.9016 0.9914
## Specificity 0.99960 0.9900 0.9055
## Pos Pred Value 0.97770 0.9662 0.9674
## Neg Pred Value 0.99744 0.9693 0.9737
## Prevalence 0.01950 0.2415 0.7390
## Detection Rate 0.01698 0.2178 0.7326
## Detection Prevalence 0.01737 0.2254 0.7573
## Balanced Accuracy 0.93523 0.9458 0.9484
## [1] "----------------------------------------------------------------------------"
## [1] "3 25 146"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13942, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8994850 0.8043598
## 2 0.8721779 0.7513073
## 3 0.8787627 0.7637221
## 4 0.8652063 0.7371751
## 5 0.8644958 0.7356377
## 6 0.8508745 0.7090613
## 7 0.8522288 0.7116464
## 8 0.8469347 0.7010256
## 9 0.8438365 0.6948077
## 10 0.8389940 0.6851783
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 271 17 29
## 1 14 7656 725
## 2 17 755 6006
##
## Overall Statistics
##
## Accuracy : 0.8995
## 95% CI : (0.8946, 0.9042)
## No Information Rate : 0.5441
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8044
##
## Mcnemar's Test P-Value : 0.2584
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89735 0.9084 0.8885
## Specificity 0.99697 0.8954 0.9116
## Pos Pred Value 0.85489 0.9120 0.8861
## Neg Pred Value 0.99796 0.8912 0.9135
## Prevalence 0.01950 0.5441 0.4364
## Detection Rate 0.01750 0.4943 0.3877
## Detection Prevalence 0.02046 0.5420 0.4376
## Balanced Accuracy 0.94716 0.9019 0.9000
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13940, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9313095 0.8651400
## 2 extratrees 0.8921224 0.7860811
## 7 gini 0.9388626 0.8803037
## 7 extratrees 0.9430592 0.8884495
## 12 gini 0.9299536 0.8627259
## 12 extratrees 0.9460290 0.8943760
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 264 2 3
## 1 11 8113 480
## 2 27 313 6277
##
## Overall Statistics
##
## Accuracy : 0.946
## 95% CI : (0.9424, 0.9495)
## No Information Rate : 0.5441
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8944
##
## Mcnemar's Test P-Value : 4.376e-13
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.87417 0.9626 0.9286
## Specificity 0.99967 0.9305 0.9611
## Pos Pred Value 0.98141 0.9429 0.9486
## Neg Pred Value 0.99750 0.9543 0.9456
## Prevalence 0.01950 0.5441 0.4364
## Detection Rate 0.01704 0.5238 0.4052
## Detection Prevalence 0.01737 0.5555 0.4272
## Balanced Accuracy 0.93692 0.9465 0.9448
## [1] "----------------------------------------------------------------------------"
## [1] "3 26 145"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13940, 13941, 13941, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9196886 0.8059030
## 2 0.8983203 0.7550486
## 3 0.9022581 0.7621636
## 4 0.8916712 0.7354080
## 5 0.8928333 0.7378336
## 6 0.8844396 0.7166678
## 7 0.8808903 0.7058660
## 8 0.8748873 0.6899098
## 9 0.8731443 0.6843440
## 10 0.8693362 0.6750099
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 270 6 37
## 1 9 3443 608
## 2 23 561 10533
##
## Overall Statistics
##
## Accuracy : 0.9197
## 95% CI : (0.9153, 0.9239)
## No Information Rate : 0.7216
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8059
##
## Mcnemar's Test P-Value : 0.1241
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89404 0.8586 0.9423
## Specificity 0.99717 0.9463 0.8646
## Pos Pred Value 0.86262 0.8480 0.9475
## Neg Pred Value 0.99789 0.9504 0.8525
## Prevalence 0.01950 0.2589 0.7216
## Detection Rate 0.01743 0.2223 0.6800
## Detection Prevalence 0.02021 0.2621 0.7177
## Balanced Accuracy 0.94560 0.9024 0.9034
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13942, 13941, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9338286 0.8312976
## 2 extratrees 0.8877993 0.6927236
## 7 gini 0.9490630 0.8735413
## 7 extratrees 0.9480958 0.8701946
## 12 gini 0.9422204 0.8565419
## 12 extratrees 0.9517104 0.8799599
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 259 2 5
## 1 2 3528 218
## 2 41 480 10955
##
## Overall Statistics
##
## Accuracy : 0.9517
## 95% CI : (0.9482, 0.955)
## No Information Rate : 0.7216
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.88
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.85762 0.8798 0.9801
## Specificity 0.99954 0.9808 0.8792
## Pos Pred Value 0.97368 0.9413 0.9546
## Neg Pred Value 0.99718 0.9590 0.9444
## Prevalence 0.01950 0.2589 0.7216
## Detection Rate 0.01672 0.2278 0.7072
## Detection Prevalence 0.01717 0.2420 0.7409
## Balanced Accuracy 0.92858 0.9303 0.9296
## [1] "----------------------------------------------------------------------------"
## [1] "3 45 126"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9223369 0.8193366
## 2 0.9039392 0.7771487
## 3 0.9085865 0.7858490
## 4 0.8986453 0.7620856
## 5 0.8961270 0.7553686
## 6 0.8865074 0.7320971
## 7 0.8865733 0.7309228
## 8 0.8805054 0.7154202
## 9 0.8785037 0.7108375
## 10 0.8735972 0.6984211
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 271 33 12
## 1 16 10279 554
## 2 15 573 3737
##
## Overall Statistics
##
## Accuracy : 0.9223
## 95% CI : (0.918, 0.9265)
## No Information Rate : 0.7027
## P-Value [Acc > NIR] : < 2e-16
##
## Kappa : 0.8193
##
## Mcnemar's Test P-Value : 0.08765
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8974 0.9443 0.8685
## Specificity 0.9970 0.8762 0.9474
## Pos Pred Value 0.8576 0.9475 0.8640
## Neg Pred Value 0.9980 0.8694 0.9493
## Prevalence 0.0195 0.7027 0.2778
## Detection Rate 0.0175 0.6636 0.2413
## Detection Prevalence 0.0204 0.7004 0.2792
## Balanced Accuracy 0.9472 0.9103 0.9080
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13942, 13940, 13942, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9372503 0.8475938
## 2 extratrees 0.8940615 0.7278701
## 7 gini 0.9522922 0.8866129
## 7 extratrees 0.9511299 0.8831424
## 12 gini 0.9471923 0.8747350
## 12 extratrees 0.9544872 0.8916002
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 258 4 3
## 1 25 10686 459
## 2 19 195 3841
##
## Overall Statistics
##
## Accuracy : 0.9545
## 95% CI : (0.9511, 0.9577)
## No Information Rate : 0.7027
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8917
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.85430 0.9817 0.8926
## Specificity 0.99954 0.8949 0.9809
## Pos Pred Value 0.97358 0.9567 0.9472
## Neg Pred Value 0.99711 0.9539 0.9596
## Prevalence 0.01950 0.7027 0.2778
## Detection Rate 0.01666 0.6899 0.2480
## Detection Prevalence 0.01711 0.7211 0.2618
## Balanced Accuracy 0.92692 0.9383 0.9368
## [1] "----------------------------------------------------------------------------"
## [1] "3 46 125"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13939, 13942, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9026474 0.8088262
## 2 0.8768248 0.7580559
## 3 0.8835396 0.7709774
## 4 0.8715977 0.7471922
## 5 0.8688867 0.7416080
## 6 0.8613334 0.7266747
## 7 0.8574583 0.7191733
## 8 0.8529389 0.7099758
## 9 0.8490652 0.7018620
## 10 0.8431270 0.6902006
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 271 19 32
## 1 13 5749 727
## 2 18 699 7962
##
## Overall Statistics
##
## Accuracy : 0.9026
## 95% CI : (0.8979, 0.9073)
## No Information Rate : 0.563
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8088
##
## Mcnemar's Test P-Value : 0.1331
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89735 0.8890 0.9130
## Specificity 0.99664 0.9180 0.8941
## Pos Pred Value 0.84161 0.8860 0.9174
## Neg Pred Value 0.99796 0.9202 0.8886
## Prevalence 0.01950 0.4175 0.5630
## Detection Rate 0.01750 0.3711 0.5140
## Detection Prevalence 0.02079 0.4189 0.5603
## Balanced Accuracy 0.94700 0.9035 0.9035
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13941, 13940, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9324078 0.8656783
## 2 extratrees 0.8884442 0.7748464
## 7 gini 0.9393807 0.8799957
## 7 extratrees 0.9453854 0.8918057
## 12 gini 0.9318279 0.8649360
## 12 extratrees 0.9488713 0.8988482
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 259 2 3
## 1 16 6006 285
## 2 27 459 8433
##
## Overall Statistics
##
## Accuracy : 0.9489
## 95% CI : (0.9453, 0.9523)
## No Information Rate : 0.563
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8989
##
## Mcnemar's Test P-Value : 2.902e-15
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.85762 0.9287 0.9670
## Specificity 0.99967 0.9666 0.9282
## Pos Pred Value 0.98106 0.9523 0.9455
## Neg Pred Value 0.99718 0.9498 0.9562
## Prevalence 0.01950 0.4175 0.5630
## Detection Rate 0.01672 0.3877 0.5444
## Detection Prevalence 0.01704 0.4072 0.5758
## Balanced Accuracy 0.92864 0.9477 0.9476
## [1] "----------------------------------------------------------------------------"
## [1] "3 56 124"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9464157 0.8707244
## 2 0.9327940 0.8376367
## 3 0.9378940 0.8482245
## 4 0.9324717 0.8343320
## 5 0.9323428 0.8329333
## 6 0.9295663 0.8255578
## 7 0.9280181 0.8213565
## 8 0.9233051 0.8091501
## 9 0.9239509 0.8100731
## 10 0.9215622 0.8039557
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 271 11 31
## 1 3 10752 366
## 2 28 391 3637
##
## Overall Statistics
##
## Accuracy : 0.9464
## 95% CI : (0.9428, 0.9499)
## No Information Rate : 0.7201
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8707
##
## Mcnemar's Test P-Value : 0.1357
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.89735 0.9640 0.9016
## Specificity 0.99723 0.9149 0.9634
## Pos Pred Value 0.86581 0.9668 0.8967
## Neg Pred Value 0.99796 0.9080 0.9653
## Prevalence 0.01950 0.7201 0.2604
## Detection Rate 0.01750 0.6941 0.2348
## Detection Prevalence 0.02021 0.7179 0.2618
## Balanced Accuracy 0.94729 0.9394 0.9325
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13941, 13939, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9546819 0.8862416
## 2 extratrees 0.9331185 0.8272936
## 7 gini 0.9668180 0.9181937
## 7 extratrees 0.9648173 0.9127952
## 12 gini 0.9632028 0.9096440
## 12 extratrees 0.9685616 0.9223843
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 259 1 6
## 1 4 11060 344
## 2 39 93 3684
##
## Overall Statistics
##
## Accuracy : 0.9686
## 95% CI : (0.9657, 0.9713)
## No Information Rate : 0.7201
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9224
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.85762 0.9916 0.9132
## Specificity 0.99954 0.9197 0.9885
## Pos Pred Value 0.97368 0.9695 0.9654
## Neg Pred Value 0.99718 0.9770 0.9700
## Prevalence 0.01950 0.7201 0.2604
## Detection Rate 0.01672 0.7140 0.2378
## Detection Prevalence 0.01717 0.7365 0.2464
## Balanced Accuracy 0.92858 0.9557 0.9509
## [1] "----------------------------------------------------------------------------"
## [1] "4 12 356"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13941, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9414467 0.8573649
## 2 0.9293093 0.8274946
## 3 0.9326021 0.8331121
## 4 0.9267277 0.8182022
## 5 0.9257592 0.8144324
## 6 0.9216276 0.8030048
## 7 0.9206594 0.7995742
## 8 0.9180770 0.7917822
## 9 0.9156239 0.7848813
## 10 0.9124603 0.7763073
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2676 32 382
## 1 45 877 44
## 2 378 26 11030
##
## Overall Statistics
##
## Accuracy : 0.9414
## 95% CI : (0.9376, 0.9451)
## No Information Rate : 0.7396
## P-Value [Acc > NIR] : < 2e-16
##
## Kappa : 0.8573
##
## Mcnemar's Test P-Value : 0.07703
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8635 0.93797 0.9628
## Specificity 0.9666 0.99389 0.8999
## Pos Pred Value 0.8660 0.90787 0.9647
## Neg Pred Value 0.9659 0.99601 0.8950
## Prevalence 0.2001 0.06036 0.7396
## Detection Rate 0.1728 0.05662 0.7121
## Detection Prevalence 0.1995 0.06236 0.7382
## Balanced Accuracy 0.9150 0.96593 0.9313
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13940, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9508711 0.8739377
## 2 extratrees 0.9249832 0.7989887
## 7 gini 0.9635255 0.9086552
## 7 extratrees 0.9652683 0.9126557
## 12 gini 0.9601688 0.9007283
## 12 extratrees 0.9666885 0.9165294
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2712 24 74
## 1 9 903 23
## 2 378 8 11359
##
## Overall Statistics
##
## Accuracy : 0.9667
## 95% CI : (0.9637, 0.9695)
## No Information Rate : 0.7396
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9166
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8751 0.96578 0.9915
## Specificity 0.9921 0.99780 0.9043
## Pos Pred Value 0.9651 0.96578 0.9671
## Neg Pred Value 0.9695 0.99780 0.9741
## Prevalence 0.2001 0.06036 0.7396
## Detection Rate 0.1751 0.05830 0.7333
## Detection Prevalence 0.1814 0.06036 0.7582
## Balanced Accuracy 0.9336 0.98179 0.9479
## [1] "----------------------------------------------------------------------------"
## [1] "4 13 256"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13940, 13942, 13940, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9420267 0.8466656
## 2 0.9260164 0.8044348
## 3 0.9320856 0.8172993
## 4 0.9248547 0.7965693
## 5 0.9250484 0.7950810
## 6 0.9200782 0.7803910
## 7 0.9191094 0.7770035
## 8 0.9152365 0.7655359
## 9 0.9131063 0.7580417
## 10 0.9139453 0.7597462
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2681 13 377
## 1 24 538 46
## 2 394 44 11373
##
## Overall Statistics
##
## Accuracy : 0.942
## 95% CI : (0.9382, 0.9457)
## No Information Rate : 0.7615
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8467
##
## Mcnemar's Test P-Value : 0.297
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8651 0.90420 0.9641
## Specificity 0.9685 0.99530 0.8814
## Pos Pred Value 0.8730 0.88487 0.9629
## Neg Pred Value 0.9663 0.99617 0.8850
## Prevalence 0.2001 0.03841 0.7615
## Detection Rate 0.1731 0.03473 0.7342
## Detection Prevalence 0.1983 0.03925 0.7625
## Balanced Accuracy 0.9168 0.94975 0.9228
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13942, 13941, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9497105 0.8586137
## 2 extratrees 0.9234998 0.7732693
## 7 gini 0.9631379 0.8992649
## 7 extratrees 0.9628791 0.8980527
## 12 gini 0.9629442 0.8994866
## 12 extratrees 0.9659777 0.9069241
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2701 12 70
## 1 8 546 10
## 2 390 37 11716
##
## Overall Statistics
##
## Accuracy : 0.966
## 95% CI : (0.963, 0.9688)
## No Information Rate : 0.7615
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.907
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8716 0.91765 0.9932
## Specificity 0.9934 0.99879 0.8844
## Pos Pred Value 0.9705 0.96809 0.9648
## Neg Pred Value 0.9687 0.99672 0.9761
## Prevalence 0.2001 0.03841 0.7615
## Detection Rate 0.1744 0.03525 0.7564
## Detection Prevalence 0.1797 0.03641 0.7839
## Balanced Accuracy 0.9325 0.95822 0.9388
## [1] "----------------------------------------------------------------------------"
## [1] "4 15 236"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13940, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8848285 0.8115880
## 2 0.8557777 0.7640789
## 3 0.8620375 0.7735608
## 4 0.8507403 0.7547630
## 5 0.8453163 0.7455258
## 6 0.8340182 0.7268256
## 7 0.8315014 0.7218969
## 8 0.8269826 0.7139728
## 9 0.8235605 0.7076964
## 10 0.8165880 0.6959301
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2693 268 135
## 1 265 7299 463
## 2 141 512 3714
##
## Overall Statistics
##
## Accuracy : 0.8848
## 95% CI : (0.8797, 0.8898)
## No Information Rate : 0.5216
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8116
##
## Mcnemar's Test P-Value : 0.4558
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8690 0.9035 0.8613
## Specificity 0.9675 0.9018 0.9416
## Pos Pred Value 0.8698 0.9093 0.8505
## Neg Pred Value 0.9672 0.8955 0.9462
## Prevalence 0.2001 0.5216 0.2784
## Detection Rate 0.1739 0.4712 0.2398
## Detection Prevalence 0.1999 0.5182 0.2819
## Balanced Accuracy 0.9182 0.9026 0.9014
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9165913 0.8609638
## 2 extratrees 0.8689475 0.7755739
## 7 gini 0.9359584 0.8944031
## 7 extratrees 0.9339576 0.8904758
## 12 gini 0.9280182 0.8813119
## 12 extratrees 0.9391223 0.8992980
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2793 63 55
## 1 227 7832 335
## 2 79 184 3922
##
## Overall Statistics
##
## Accuracy : 0.9391
## 95% CI : (0.9352, 0.9428)
## No Information Rate : 0.5216
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8993
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9013 0.9694 0.9096
## Specificity 0.9905 0.9242 0.9765
## Pos Pred Value 0.9595 0.9330 0.9372
## Neg Pred Value 0.9757 0.9652 0.9655
## Prevalence 0.2001 0.5216 0.2784
## Detection Rate 0.1803 0.5056 0.2532
## Detection Prevalence 0.1879 0.5419 0.2702
## Balanced Accuracy 0.9459 0.9468 0.9430
## [1] "----------------------------------------------------------------------------"
## [1] "4 16 235"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8884436 0.8099499
## 2 0.8633313 0.7672010
## 3 0.8653316 0.7689716
## 4 0.8533890 0.7483733
## 5 0.8510652 0.7438586
## 6 0.8431245 0.7294258
## 7 0.8386700 0.7213997
## 8 0.8324084 0.7097083
## 9 0.8316984 0.7075584
## 10 0.8250487 0.6955164
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2696 86 311
## 1 90 3124 477
## 2 313 451 7942
##
## Overall Statistics
##
## Accuracy : 0.8884
## 95% CI : (0.8834, 0.8934)
## No Information Rate : 0.5636
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8099
##
## Mcnemar's Test P-Value : 0.8433
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8700 0.8533 0.9097
## Specificity 0.9680 0.9521 0.8870
## Pos Pred Value 0.8716 0.8464 0.9122
## Neg Pred Value 0.9675 0.9545 0.8838
## Prevalence 0.2001 0.2363 0.5636
## Detection Rate 0.1740 0.2017 0.5127
## Detection Prevalence 0.1997 0.2383 0.5620
## Balanced Accuracy 0.9190 0.9027 0.8984
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9160739 0.8525646
## 2 extratrees 0.8577801 0.7396682
## 7 gini 0.9364101 0.8902274
## 7 extratrees 0.9352477 0.8874908
## 12 gini 0.9262745 0.8726723
## 12 extratrees 0.9402188 0.8965124
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2759 27 67
## 1 62 3304 162
## 2 278 330 8501
##
## Overall Statistics
##
## Accuracy : 0.9402
## 95% CI : (0.9364, 0.9439)
## No Information Rate : 0.5636
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8966
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8903 0.9025 0.9738
## Specificity 0.9924 0.9811 0.9101
## Pos Pred Value 0.9671 0.9365 0.9333
## Neg Pred Value 0.9731 0.9702 0.9641
## Prevalence 0.2001 0.2363 0.5636
## Detection Rate 0.1781 0.2133 0.5488
## Detection Prevalence 0.1842 0.2278 0.5881
## Balanced Accuracy 0.9414 0.9418 0.9419
## [1] "----------------------------------------------------------------------------"
## [1] "4 23 156"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13942, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9391212 0.8522201
## 2 0.9233025 0.8141923
## 3 0.9282747 0.8231126
## 4 0.9209796 0.8049893
## 5 0.9225293 0.8071310
## 6 0.9176230 0.7946394
## 7 0.9166551 0.7908441
## 8 0.9132990 0.7819531
## 9 0.9120716 0.7780009
## 10 0.9085854 0.7683827
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2689 40 367
## 1 59 853 75
## 2 351 51 11005
##
## Overall Statistics
##
## Accuracy : 0.9391
## 95% CI : (0.9352, 0.9428)
## No Information Rate : 0.739
## P-Value [Acc > NIR] : < 2e-16
##
## Kappa : 0.8522
##
## Mcnemar's Test P-Value : 0.03552
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8677 0.90360 0.9614
## Specificity 0.9672 0.99079 0.9006
## Pos Pred Value 0.8685 0.86424 0.9648
## Neg Pred Value 0.9669 0.99373 0.8917
## Prevalence 0.2001 0.06094 0.7390
## Detection Rate 0.1736 0.05507 0.7105
## Detection Prevalence 0.1999 0.06372 0.7364
## Balanced Accuracy 0.9174 0.94719 0.9310
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13942, 13940, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9513244 0.8755860
## 2 extratrees 0.9219505 0.7911196
## 7 gini 0.9643647 0.9110199
## 7 extratrees 0.9639782 0.9094582
## 12 gini 0.9617826 0.9048972
## 12 extratrees 0.9667538 0.9167968
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2724 27 61
## 1 13 896 31
## 2 362 21 11355
##
## Overall Statistics
##
## Accuracy : 0.9668
## 95% CI : (0.9638, 0.9695)
## No Information Rate : 0.739
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9169
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8790 0.94915 0.9920
## Specificity 0.9929 0.99698 0.9053
## Pos Pred Value 0.9687 0.95319 0.9674
## Neg Pred Value 0.9704 0.99670 0.9755
## Prevalence 0.2001 0.06094 0.7390
## Detection Rate 0.1759 0.05784 0.7331
## Detection Prevalence 0.1815 0.06068 0.7578
## Balanced Accuracy 0.9359 0.97306 0.9486
## [1] "----------------------------------------------------------------------------"
## [1] "4 25 136"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8854089 0.8085817
## 2 0.8578434 0.7626725
## 3 0.8617170 0.7677009
## 4 0.8488055 0.7461572
## 5 0.8480954 0.7439754
## 6 0.8400908 0.7304217
## 7 0.8361525 0.7232194
## 8 0.8302773 0.7124174
## 9 0.8256928 0.7039736
## 10 0.8224652 0.6980393
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2695 310 106
## 1 294 7646 483
## 2 110 472 3374
##
## Overall Statistics
##
## Accuracy : 0.8854
## 95% CI : (0.8803, 0.8904)
## No Information Rate : 0.5441
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8086
##
## Mcnemar's Test P-Value : 0.8908
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8696 0.9072 0.8514
## Specificity 0.9664 0.8900 0.9495
## Pos Pred Value 0.8663 0.9078 0.8529
## Neg Pred Value 0.9674 0.8893 0.9489
## Prevalence 0.2001 0.5441 0.2558
## Detection Rate 0.1740 0.4936 0.2178
## Detection Prevalence 0.2008 0.5438 0.2554
## Balanced Accuracy 0.9180 0.8986 0.9004
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13940, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9187216 0.8610777
## 2 extratrees 0.8644292 0.7599647
## 7 gini 0.9356360 0.8915584
## 7 extratrees 0.9343450 0.8885964
## 12 gini 0.9262104 0.8757152
## 12 extratrees 0.9398969 0.8983798
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2785 66 33
## 1 243 8182 338
## 2 71 180 3592
##
## Overall Statistics
##
## Accuracy : 0.9399
## 95% CI : (0.936, 0.9436)
## No Information Rate : 0.5441
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8984
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8987 0.9708 0.9064
## Specificity 0.9920 0.9177 0.9782
## Pos Pred Value 0.9657 0.9337 0.9347
## Neg Pred Value 0.9751 0.9634 0.9681
## Prevalence 0.2001 0.5441 0.2558
## Detection Rate 0.1798 0.5282 0.2319
## Detection Prevalence 0.1862 0.5657 0.2481
## Balanced Accuracy 0.9453 0.9443 0.9423
## [1] "----------------------------------------------------------------------------"
## [1] "4 26 135"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8870244 0.8120247
## 2 0.8556498 0.7602799
## 3 0.8648173 0.7741288
## 4 0.8496453 0.7486847
## 5 0.8495814 0.7480899
## 6 0.8413170 0.7336835
## 7 0.8375088 0.7266375
## 8 0.8288579 0.7114352
## 9 0.8264696 0.7066365
## 10 0.8233068 0.7011643
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2700 123 291
## 1 120 3434 484
## 2 279 453 7606
##
## Overall Statistics
##
## Accuracy : 0.887
## 95% CI : (0.8819, 0.892)
## No Information Rate : 0.5411
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.812
##
## Mcnemar's Test P-Value : 0.7255
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8712 0.8564 0.9075
## Specificity 0.9666 0.9474 0.8970
## Pos Pred Value 0.8671 0.8504 0.9122
## Neg Pred Value 0.9678 0.9497 0.8916
## Prevalence 0.2001 0.2589 0.5411
## Detection Rate 0.1743 0.2217 0.4910
## Detection Prevalence 0.2010 0.2607 0.5383
## Balanced Accuracy 0.9189 0.9019 0.9023
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9179489 0.8602226
## 2 extratrees 0.8655916 0.7629753
## 7 gini 0.9336345 0.8884435
## 7 extratrees 0.9353138 0.8905430
## 12 gini 0.9279532 0.8789290
## 12 extratrees 0.9402197 0.8991753
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2776 48 64
## 1 67 3638 167
## 2 256 324 8150
##
## Overall Statistics
##
## Accuracy : 0.9402
## 95% CI : (0.9364, 0.9439)
## No Information Rate : 0.5411
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8992
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8958 0.9072 0.9724
## Specificity 0.9910 0.9796 0.9184
## Pos Pred Value 0.9612 0.9396 0.9336
## Neg Pred Value 0.9744 0.9680 0.9658
## Prevalence 0.2001 0.2589 0.5411
## Detection Rate 0.1792 0.2349 0.5261
## Detection Prevalence 0.1864 0.2500 0.5636
## Balanced Accuracy 0.9434 0.9434 0.9454
## [1] "----------------------------------------------------------------------------"
## [1] "4 35 126"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13942, 13940, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8900591 0.8199845
## 2 0.8617183 0.7735159
## 3 0.8686926 0.7841510
## 4 0.8533926 0.7585797
## 5 0.8516509 0.7553739
## 6 0.8429344 0.7407288
## 7 0.8397714 0.7350410
## 8 0.8331865 0.7237734
## 9 0.8309907 0.7197181
## 10 0.8257611 0.7103586
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2702 292 116
## 1 273 7340 442
## 2 124 456 3745
##
## Overall Statistics
##
## Accuracy : 0.8901
## 95% CI : (0.885, 0.8949)
## No Information Rate : 0.5221
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.82
##
## Mcnemar's Test P-Value : 0.7713
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8719 0.9075 0.8703
## Specificity 0.9671 0.9034 0.9482
## Pos Pred Value 0.8688 0.9112 0.8659
## Neg Pred Value 0.9679 0.8994 0.9500
## Prevalence 0.2001 0.5221 0.2778
## Detection Rate 0.1744 0.4739 0.2418
## Detection Prevalence 0.2008 0.5200 0.2792
## Balanced Accuracy 0.9195 0.9055 0.9092
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13940, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9188521 0.8643072
## 2 extratrees 0.8699818 0.7766410
## 7 gini 0.9380893 0.8978472
## 7 extratrees 0.9363465 0.8943557
## 12 gini 0.9324723 0.8885426
## 12 extratrees 0.9397685 0.9003212
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2782 63 51
## 1 246 7836 313
## 2 71 189 3939
##
## Overall Statistics
##
## Accuracy : 0.9398
## 95% CI : (0.9359, 0.9435)
## No Information Rate : 0.5221
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9003
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8977 0.9688 0.9154
## Specificity 0.9908 0.9245 0.9768
## Pos Pred Value 0.9606 0.9334 0.9381
## Neg Pred Value 0.9748 0.9645 0.9678
## Prevalence 0.2001 0.5221 0.2778
## Detection Rate 0.1796 0.5059 0.2543
## Detection Prevalence 0.1870 0.5420 0.2711
## Balanced Accuracy 0.9443 0.9467 0.9461
## [1] "----------------------------------------------------------------------------"
## [1] "4 36 125"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8901878 0.8129410
## 2 0.8630739 0.7667930
## 3 0.8668182 0.7719780
## 4 0.8538413 0.7496552
## 5 0.8537765 0.7489627
## 6 0.8455767 0.7339677
## 7 0.8400254 0.7242351
## 8 0.8343443 0.7138849
## 9 0.8327296 0.7102699
## 10 0.8280170 0.7018771
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2691 96 307
## 1 109 3130 446
## 2 299 444 7968
##
## Overall Statistics
##
## Accuracy : 0.8902
## 95% CI : (0.8852, 0.8951)
## No Information Rate : 0.563
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8129
##
## Mcnemar's Test P-Value : 0.8171
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8683 0.8529 0.9137
## Specificity 0.9675 0.9530 0.8902
## Pos Pred Value 0.8697 0.8494 0.9147
## Neg Pred Value 0.9671 0.9543 0.8889
## Prevalence 0.2001 0.2369 0.5630
## Detection Rate 0.1737 0.2021 0.5144
## Detection Prevalence 0.1997 0.2379 0.5624
## Balanced Accuracy 0.9179 0.9030 0.9019
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9171726 0.8548371
## 2 extratrees 0.8620399 0.7485745
## 7 gini 0.9347961 0.8877232
## 7 extratrees 0.9371847 0.8910802
## 12 gini 0.9251767 0.8711720
## 12 extratrees 0.9393152 0.8950852
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2784 41 69
## 1 63 3277 163
## 2 252 352 8489
##
## Overall Statistics
##
## Accuracy : 0.9393
## 95% CI : (0.9354, 0.943)
## No Information Rate : 0.563
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8951
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8984 0.8929 0.9734
## Specificity 0.9911 0.9809 0.9108
## Pos Pred Value 0.9620 0.9355 0.9336
## Neg Pred Value 0.9750 0.9672 0.9637
## Prevalence 0.2001 0.2369 0.5630
## Detection Rate 0.1797 0.2116 0.5480
## Detection Prevalence 0.1868 0.2261 0.5870
## Balanced Accuracy 0.9447 0.9369 0.9421
## [1] "----------------------------------------------------------------------------"
## [1] "4 56 123"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13940, 13941, 13941, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9433825 0.8702188
## 2 0.9311164 0.8417899
## 3 0.9348600 0.8483590
## 4 0.9284037 0.8335170
## 5 0.9284046 0.8319067
## 6 0.9242723 0.8217650
## 7 0.9229167 0.8179951
## 8 0.9198191 0.8097360
## 9 0.9193662 0.8083484
## 10 0.9169778 0.8019186
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2681 362 40
## 1 356 10754 19
## 2 62 38 1178
##
## Overall Statistics
##
## Accuracy : 0.9434
## 95% CI : (0.9396, 0.947)
## No Information Rate : 0.7201
## P-Value [Acc > NIR] : < 2e-16
##
## Kappa : 0.8702
##
## Mcnemar's Test P-Value : 0.01105
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8651 0.9641 0.95230
## Specificity 0.9676 0.9135 0.99298
## Pos Pred Value 0.8696 0.9663 0.92175
## Neg Pred Value 0.9663 0.9083 0.99585
## Prevalence 0.2001 0.7201 0.07986
## Detection Rate 0.1731 0.6943 0.07605
## Detection Prevalence 0.1990 0.7185 0.08250
## Balanced Accuracy 0.9163 0.9388 0.97264
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13942, 13941, 13942, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9528738 0.8868377
## 2 extratrees 0.9255651 0.8146344
## 7 gini 0.9668178 0.9221753
## 7 extratrees 0.9662363 0.9204611
## 12 gini 0.9622337 0.9114808
## 12 extratrees 0.9692061 0.9277158
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2740 63 28
## 1 341 11072 8
## 2 18 19 1201
##
## Overall Statistics
##
## Accuracy : 0.9692
## 95% CI : (0.9664, 0.9719)
## No Information Rate : 0.7201
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9277
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8842 0.9926 0.97090
## Specificity 0.9927 0.9195 0.99740
## Pos Pred Value 0.9679 0.9694 0.97011
## Neg Pred Value 0.9716 0.9798 0.99747
## Prevalence 0.2001 0.7201 0.07986
## Detection Rate 0.1769 0.7148 0.07753
## Detection Prevalence 0.1828 0.7373 0.07992
## Balanced Accuracy 0.9384 0.9561 0.98415
## [1] "----------------------------------------------------------------------------"
## [1] "5 12 346"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13942, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9002600 0.8198694
## 2 0.8726926 0.7701433
## 3 0.8818616 0.7863763
## 4 0.8671416 0.7598726
## 5 0.8666253 0.7587737
## 6 0.8568117 0.7409493
## 7 0.8541002 0.7358157
## 8 0.8466760 0.7223355
## 9 0.8464823 0.7218290
## 10 0.8399622 0.7099018
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7028 13 667
## 1 23 876 61
## 2 735 46 6041
##
## Overall Statistics
##
## Accuracy : 0.9003
## 95% CI : (0.8954, 0.9049)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2e-16
##
## Kappa : 0.8199
##
## Mcnemar's Test P-Value : 0.04246
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9026 0.93690 0.8925
## Specificity 0.9117 0.99423 0.9104
## Pos Pred Value 0.9118 0.91250 0.8855
## Neg Pred Value 0.9026 0.99594 0.9160
## Prevalence 0.5026 0.06036 0.4370
## Detection Rate 0.4537 0.05655 0.3900
## Detection Prevalence 0.4976 0.06198 0.4404
## Balanced Accuracy 0.9072 0.96556 0.9014
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13940, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9317620 0.8760855
## 2 extratrees 0.8919954 0.8023572
## 7 gini 0.9445444 0.8995499
## 7 extratrees 0.9451901 0.9006813
## 12 gini 0.9399616 0.8912619
## 12 extratrees 0.9492569 0.9080788
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7487 2 437
## 1 7 904 19
## 2 292 29 6313
##
## Overall Statistics
##
## Accuracy : 0.9493
## 95% CI : (0.9457, 0.9527)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9081
##
## Mcnemar's Test P-Value : 2.29e-07
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9616 0.96684 0.9326
## Specificity 0.9430 0.99821 0.9632
## Pos Pred Value 0.9446 0.97204 0.9516
## Neg Pred Value 0.9605 0.99787 0.9485
## Prevalence 0.5026 0.06036 0.4370
## Detection Rate 0.4833 0.05836 0.4076
## Detection Prevalence 0.5117 0.06004 0.4283
## Balanced Accuracy 0.9523 0.98253 0.9479
## [1] "----------------------------------------------------------------------------"
## [1] "5 13 246"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13941, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9017430 0.8165942
## 2 0.8757915 0.7680851
## 3 0.8816661 0.7788319
## 4 0.8691416 0.7552711
## 5 0.8704984 0.7578138
## 6 0.8604907 0.7389969
## 7 0.8570047 0.7320900
## 8 0.8491932 0.7175246
## 9 0.8491930 0.7173724
## 10 0.8435117 0.7067708
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7056 8 685
## 1 18 539 51
## 2 712 48 6373
##
## Overall Statistics
##
## Accuracy : 0.9017
## 95% CI : (0.8969, 0.9064)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8166
##
## Mcnemar's Test P-Value : 0.216
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9062 0.90588 0.8965
## Specificity 0.9100 0.99537 0.9093
## Pos Pred Value 0.9106 0.88651 0.8935
## Neg Pred Value 0.9057 0.99624 0.9119
## Prevalence 0.5026 0.03841 0.4589
## Detection Rate 0.4555 0.03480 0.4114
## Detection Prevalence 0.5003 0.03925 0.4605
## Balanced Accuracy 0.9081 0.95062 0.9029
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13942, 13941, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9345388 0.8771292
## 2 extratrees 0.8970303 0.8057569
## 7 gini 0.9446757 0.8963200
## 7 extratrees 0.9464822 0.8997833
## 12 gini 0.9388017 0.8852852
## 12 extratrees 0.9508730 0.9079978
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7487 1 401
## 1 1 547 13
## 2 298 47 6695
##
## Overall Statistics
##
## Accuracy : 0.9509
## 95% CI : (0.9473, 0.9542)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.908
##
## Mcnemar's Test P-Value : 1.596e-07
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9616 0.91933 0.9418
## Specificity 0.9478 0.99906 0.9588
## Pos Pred Value 0.9490 0.97504 0.9510
## Neg Pred Value 0.9607 0.99678 0.9510
## Prevalence 0.5026 0.03841 0.4589
## Detection Rate 0.4833 0.03531 0.4322
## Detection Prevalence 0.5093 0.03622 0.4545
## Balanced Accuracy 0.9547 0.95919 0.9503
## [1] "----------------------------------------------------------------------------"
## [1] "5 14 236"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8854713 0.8159912
## 2 0.8570023 0.7702190
## 3 0.8630712 0.7790211
## 4 0.8508059 0.7591731
## 5 0.8486120 0.7553145
## 6 0.8366037 0.7355950
## 7 0.8347958 0.7323602
## 8 0.8278224 0.7207155
## 9 0.8240793 0.7141761
## 10 0.8190440 0.7058115
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7050 267 432
## 1 276 2947 161
## 2 460 178 3719
##
## Overall Statistics
##
## Accuracy : 0.8855
## 95% CI : (0.8804, 0.8904)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.816
##
## Mcnemar's Test P-Value : 0.5976
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9055 0.8688 0.8625
## Specificity 0.9093 0.9639 0.9429
## Pos Pred Value 0.9098 0.8709 0.8536
## Neg Pred Value 0.9049 0.9632 0.9467
## Prevalence 0.5026 0.2190 0.2784
## Detection Rate 0.4551 0.1903 0.2401
## Detection Prevalence 0.5003 0.2185 0.2813
## Balanced Accuracy 0.9074 0.9163 0.9027
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13940, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9192384 0.8681244
## 2 extratrees 0.8696569 0.7821734
## 7 gini 0.9375740 0.8989908
## 7 extratrees 0.9352494 0.8948228
## 12 gini 0.9295689 0.8860622
## 12 extratrees 0.9396391 0.9022188
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7542 220 300
## 1 64 3075 74
## 2 180 97 3938
##
## Overall Statistics
##
## Accuracy : 0.9396
## 95% CI : (0.9358, 0.9433)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9022
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9687 0.9065 0.9133
## Specificity 0.9325 0.9886 0.9752
## Pos Pred Value 0.9355 0.9570 0.9343
## Neg Pred Value 0.9672 0.9742 0.9668
## Prevalence 0.5026 0.2190 0.2784
## Detection Rate 0.4869 0.1985 0.2542
## Detection Prevalence 0.5205 0.2074 0.2721
## Balanced Accuracy 0.9506 0.9476 0.9442
## [1] "----------------------------------------------------------------------------"
## [1] "5 16 234"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13941, 13940, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8912199 0.8257476
## 2 0.8608782 0.7774064
## 3 0.8707551 0.7922792
## 4 0.8563585 0.7691781
## 5 0.8566172 0.7691578
## 6 0.8481598 0.7553736
## 7 0.8459650 0.7513025
## 8 0.8405431 0.7425918
## 9 0.8351844 0.7333674
## 10 0.8295677 0.7241999
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7040 414 275
## 1 435 3124 127
## 2 311 123 3641
##
## Overall Statistics
##
## Accuracy : 0.8912
## 95% CI : (0.8862, 0.8961)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8258
##
## Mcnemar's Test P-Value : 0.4243
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9042 0.8533 0.9006
## Specificity 0.9106 0.9525 0.9621
## Pos Pred Value 0.9109 0.8475 0.8935
## Neg Pred Value 0.9039 0.9545 0.9648
## Prevalence 0.5026 0.2363 0.2610
## Detection Rate 0.4545 0.2017 0.2351
## Detection Prevalence 0.4990 0.2380 0.2631
## Balanced Accuracy 0.9074 0.9029 0.9313
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13942, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9204018 0.8704204
## 2 extratrees 0.8729531 0.7885933
## 7 gini 0.9419640 0.9065266
## 7 extratrees 0.9383484 0.9001997
## 12 gini 0.9338943 0.8936942
## 12 extratrees 0.9418341 0.9060083
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 7, splitrule = gini
## and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7483 271 186
## 1 190 3328 77
## 2 113 62 3780
##
## Overall Statistics
##
## Accuracy : 0.942
## 95% CI : (0.9382, 0.9456)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9065
##
## Mcnemar's Test P-Value : 2.322e-07
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9611 0.9090 0.9349
## Specificity 0.9407 0.9774 0.9847
## Pos Pred Value 0.9424 0.9257 0.9558
## Neg Pred Value 0.9599 0.9720 0.9772
## Prevalence 0.5026 0.2363 0.2610
## Detection Rate 0.4831 0.2148 0.2440
## Detection Prevalence 0.5126 0.2321 0.2553
## Balanced Accuracy 0.9509 0.9432 0.9598
## [1] "----------------------------------------------------------------------------"
## [1] "5 23 146"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13940, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8976746 0.8155386
## 2 0.8709470 0.7670820
## 3 0.8786949 0.7812038
## 4 0.8633956 0.7534767
## 5 0.8628141 0.7525732
## 6 0.8499682 0.7294496
## 7 0.8498381 0.7288868
## 8 0.8413801 0.7136013
## 9 0.8400887 0.7113805
## 10 0.8336980 0.6995271
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7045 13 661
## 1 32 862 101
## 2 709 69 5998
##
## Overall Statistics
##
## Accuracy : 0.8977
## 95% CI : (0.8928, 0.9024)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2e-16
##
## Kappa : 0.8155
##
## Mcnemar's Test P-Value : 0.00129
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9048 0.91314 0.8873
## Specificity 0.9125 0.99086 0.9109
## Pos Pred Value 0.9127 0.86633 0.8852
## Neg Pred Value 0.9046 0.99434 0.9126
## Prevalence 0.5026 0.06094 0.4364
## Detection Rate 0.4548 0.05565 0.3872
## Detection Prevalence 0.4983 0.06423 0.4374
## Balanced Accuracy 0.9087 0.95200 0.8991
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13941, 13940, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9308586 0.8744315
## 2 extratrees 0.8894133 0.7976019
## 7 gini 0.9443501 0.8991571
## 7 extratrees 0.9444796 0.8993996
## 12 gini 0.9365389 0.8849544
## 12 extratrees 0.9489986 0.9076454
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7495 4 419
## 1 7 893 29
## 2 284 47 6312
##
## Overall Statistics
##
## Accuracy : 0.949
## 95% CI : (0.9454, 0.9524)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9076
##
## Mcnemar's Test P-Value : 8.476e-07
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9626 0.94597 0.9337
## Specificity 0.9451 0.99753 0.9621
## Pos Pred Value 0.9466 0.96125 0.9502
## Neg Pred Value 0.9616 0.99650 0.9494
## Prevalence 0.5026 0.06094 0.4364
## Detection Rate 0.4839 0.05765 0.4075
## Detection Prevalence 0.5112 0.05997 0.4289
## Balanced Accuracy 0.9539 0.97175 0.9479
## [1] "----------------------------------------------------------------------------"
## [1] "5 24 136"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13941, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8874765 0.8196752
## 2 0.8617169 0.7786708
## 3 0.8639116 0.7812637
## 4 0.8542919 0.7655823
## 5 0.8526130 0.7626963
## 6 0.8444786 0.7492614
## 7 0.8414447 0.7440602
## 8 0.8355054 0.7341129
## 9 0.8346015 0.7322267
## 10 0.8268551 0.7194470
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7042 291 423
## 1 303 3304 139
## 2 441 146 3401
##
## Overall Statistics
##
## Accuracy : 0.8875
## 95% CI : (0.8824, 0.8924)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8197
##
## Mcnemar's Test P-Value : 0.852
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9044 0.8832 0.8582
## Specificity 0.9073 0.9624 0.9491
## Pos Pred Value 0.9079 0.8820 0.8528
## Neg Pred Value 0.9038 0.9628 0.9511
## Prevalence 0.5026 0.2415 0.2558
## Detection Rate 0.4546 0.2133 0.2196
## Detection Prevalence 0.5007 0.2418 0.2575
## Balanced Accuracy 0.9059 0.9228 0.9036
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13941, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9182708 0.8669221
## 2 extratrees 0.8732735 0.7892090
## 7 gini 0.9395746 0.9026520
## 7 extratrees 0.9373794 0.8985952
## 12 gini 0.9346680 0.8948236
## 12 extratrees 0.9420921 0.9064309
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7548 214 306
## 1 70 3450 62
## 2 168 77 3595
##
## Overall Statistics
##
## Accuracy : 0.9421
## 95% CI : (0.9383, 0.9457)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9064
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9694 0.9222 0.9071
## Specificity 0.9325 0.9888 0.9787
## Pos Pred Value 0.9355 0.9631 0.9362
## Neg Pred Value 0.9679 0.9756 0.9684
## Prevalence 0.5026 0.2415 0.2558
## Detection Rate 0.4873 0.2227 0.2321
## Detection Prevalence 0.5209 0.2312 0.2479
## Balanced Accuracy 0.9510 0.9555 0.9429
## [1] "----------------------------------------------------------------------------"
## [1] "5 26 134"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13942, 13941, 13942, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8888952 0.8219449
## 2 0.8604907 0.7767874
## 3 0.8650741 0.7832258
## 4 0.8506134 0.7597141
## 5 0.8501617 0.7587817
## 6 0.8400256 0.7421370
## 7 0.8375727 0.7377619
## 8 0.8313106 0.7269755
## 9 0.8285997 0.7225078
## 10 0.8234996 0.7138932
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7069 427 258
## 1 438 3429 165
## 2 279 154 3271
##
## Overall Statistics
##
## Accuracy : 0.8889
## 95% CI : (0.8838, 0.8938)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8219
##
## Mcnemar's Test P-Value : 0.7196
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9079 0.8551 0.8855
## Specificity 0.9111 0.9475 0.9633
## Pos Pred Value 0.9117 0.8504 0.8831
## Neg Pred Value 0.9073 0.9493 0.9641
## Prevalence 0.5026 0.2589 0.2385
## Detection Rate 0.4564 0.2214 0.2112
## Detection Prevalence 0.5006 0.2603 0.2391
## Balanced Accuracy 0.9095 0.9013 0.9244
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13940, 13942, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9197553 0.8693838
## 2 extratrees 0.8720468 0.7869272
## 7 gini 0.9384122 0.9007589
## 7 extratrees 0.9358944 0.8961447
## 12 gini 0.9337636 0.8932236
## 12 extratrees 0.9398333 0.9027896
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7530 309 218
## 1 180 3635 83
## 2 76 66 3393
##
## Overall Statistics
##
## Accuracy : 0.9398
## 95% CI : (0.936, 0.9435)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9028
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9671 0.9065 0.9185
## Specificity 0.9316 0.9771 0.9880
## Pos Pred Value 0.9346 0.9325 0.9598
## Neg Pred Value 0.9656 0.9677 0.9748
## Prevalence 0.5026 0.2589 0.2385
## Detection Rate 0.4861 0.2347 0.2190
## Detection Prevalence 0.5201 0.2516 0.2282
## Balanced Accuracy 0.9494 0.9418 0.9532
## [1] "----------------------------------------------------------------------------"
## [1] "5 34 126"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13941, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8896070 0.8227354
## 2 0.8624934 0.7790768
## 3 0.8675938 0.7866738
## 4 0.8526160 0.7623893
## 5 0.8543585 0.7648919
## 6 0.8450618 0.7494259
## 7 0.8423495 0.7447150
## 8 0.8370560 0.7357556
## 9 0.8320861 0.7274375
## 10 0.8270502 0.7189195
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7046 264 424
## 1 280 2991 136
## 2 460 146 3743
##
## Overall Statistics
##
## Accuracy : 0.8896
## 95% CI : (0.8846, 0.8945)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8227
##
## Mcnemar's Test P-Value : 0.5142
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9050 0.8794 0.8699
## Specificity 0.9107 0.9656 0.9458
## Pos Pred Value 0.9110 0.8779 0.8607
## Neg Pred Value 0.9046 0.9661 0.9497
## Prevalence 0.5026 0.2196 0.2778
## Detection Rate 0.4549 0.1931 0.2416
## Detection Prevalence 0.4993 0.2199 0.2808
## Balanced Accuracy 0.9078 0.9225 0.9078
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13942, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9196898 0.8689003
## 2 extratrees 0.8739819 0.7895543
## 7 gini 0.9402836 0.9035029
## 7 extratrees 0.9364749 0.8968955
## 12 gini 0.9345376 0.8941530
## 12 extratrees 0.9410586 0.9045364
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7527 218 307
## 1 76 3104 50
## 2 183 79 3946
##
## Overall Statistics
##
## Accuracy : 0.9411
## 95% CI : (0.9372, 0.9447)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9045
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9667 0.9127 0.9170
## Specificity 0.9319 0.9896 0.9766
## Pos Pred Value 0.9348 0.9610 0.9377
## Neg Pred Value 0.9652 0.9758 0.9684
## Prevalence 0.5026 0.2196 0.2778
## Detection Rate 0.4859 0.2004 0.2547
## Detection Prevalence 0.5198 0.2085 0.2717
## Balanced Accuracy 0.9493 0.9511 0.9468
## [1] "----------------------------------------------------------------------------"
## [1] "5 36 124"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13942, 13940, 13941, 13942, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8919318 0.8268242
## 2 0.8651396 0.7839653
## 3 0.8706926 0.7923900
## 4 0.8588773 0.7731706
## 5 0.8575879 0.7709437
## 6 0.8499059 0.7584622
## 7 0.8459027 0.7516428
## 8 0.8355087 0.7343053
## 9 0.8349929 0.7333028
## 10 0.8296985 0.7242835
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7057 413 280
## 1 443 3132 127
## 2 286 125 3627
##
## Overall Statistics
##
## Accuracy : 0.8919
## 95% CI : (0.8869, 0.8968)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8268
##
## Mcnemar's Test P-Value : 0.7696
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9064 0.8534 0.8991
## Specificity 0.9100 0.9518 0.9641
## Pos Pred Value 0.9106 0.8460 0.8982
## Neg Pred Value 0.9058 0.9544 0.9645
## Prevalence 0.5026 0.2369 0.2604
## Detection Rate 0.4556 0.2022 0.2342
## Detection Prevalence 0.5003 0.2390 0.2607
## Balanced Accuracy 0.9082 0.9026 0.9316
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13940, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9196279 0.8689771
## 2 extratrees 0.8714028 0.7857420
## 7 gini 0.9382196 0.9004051
## 7 extratrees 0.9370576 0.8980155
## 12 gini 0.9307315 0.8883792
## 12 extratrees 0.9412537 0.9050390
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7537 309 216
## 1 161 3290 65
## 2 88 71 3753
##
## Overall Statistics
##
## Accuracy : 0.9413
## 95% CI : (0.9374, 0.9449)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9051
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9680 0.8965 0.9303
## Specificity 0.9319 0.9809 0.9861
## Pos Pred Value 0.9349 0.9357 0.9594
## Neg Pred Value 0.9665 0.9683 0.9757
## Prevalence 0.5026 0.2369 0.2604
## Detection Rate 0.4866 0.2124 0.2423
## Detection Prevalence 0.5205 0.2270 0.2526
## Balanced Accuracy 0.9499 0.9387 0.9582
## [1] "----------------------------------------------------------------------------"
## [1] "5 46 123"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13942, 13942, 13940, 13940, 13943, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8992242 0.8225641
## 2 0.8772115 0.7836850
## 3 0.8808921 0.7900751
## 4 0.8700463 0.7708514
## 5 0.8684326 0.7678741
## 6 0.8566180 0.7472597
## 7 0.8553916 0.7449067
## 8 0.8484859 0.7327691
## 9 0.8441610 0.7248633
## 10 0.8406102 0.7186907
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7027 674 19
## 1 721 5727 43
## 2 38 66 1175
##
## Overall Statistics
##
## Accuracy : 0.8992
## 95% CI : (0.8944, 0.9039)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8226
##
## Mcnemar's Test P-Value : 0.005161
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9025 0.8856 0.94988
## Specificity 0.9100 0.9153 0.99270
## Pos Pred Value 0.9102 0.8823 0.91869
## Neg Pred Value 0.9023 0.9178 0.99564
## Prevalence 0.5026 0.4175 0.07986
## Detection Rate 0.4536 0.3697 0.07586
## Detection Prevalence 0.4984 0.4190 0.08257
## Balanced Accuracy 0.9063 0.9005 0.97129
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13940, 13942, 13940, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9340855 0.8831554
## 2 extratrees 0.8918007 0.8066469
## 7 gini 0.9442859 0.9014470
## 7 extratrees 0.9468029 0.9059226
## 12 gini 0.9368624 0.8882469
## 12 extratrees 0.9502895 0.9121017
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 7497 427 2
## 1 277 6027 39
## 2 12 13 1196
##
## Overall Statistics
##
## Accuracy : 0.9503
## 95% CI : (0.9468, 0.9537)
## No Information Rate : 0.5026
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9121
##
## Mcnemar's Test P-Value : 2.847e-11
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9629 0.9320 0.96686
## Specificity 0.9443 0.9650 0.99825
## Pos Pred Value 0.9459 0.9502 0.97952
## Neg Pred Value 0.9618 0.9519 0.99713
## Prevalence 0.5026 0.4175 0.07986
## Detection Rate 0.4840 0.3891 0.07721
## Detection Prevalence 0.5117 0.4095 0.07883
## Balanced Accuracy 0.9536 0.9485 0.98255
## [1] "----------------------------------------------------------------------------"
## [1] "6 12 345"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13942, 13941, 13941, 13940, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9240805 0.8227434
## 2 0.9041320 0.7771365
## 3 0.9117506 0.7921310
## 4 0.8975466 0.7586442
## 5 0.8998711 0.7630839
## 6 0.8911551 0.7421746
## 7 0.8901222 0.7378664
## 8 0.8859255 0.7274539
## 9 0.8834727 0.7199601
## 10 0.8778557 0.7053701
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2852 2 519
## 1 4 871 77
## 2 512 62 10591
##
## Overall Statistics
##
## Accuracy : 0.9241
## 95% CI : (0.9198, 0.9282)
## No Information Rate : 0.7222
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8228
##
## Mcnemar's Test P-Value : 0.5062
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8468 0.93155 0.9467
## Specificity 0.9570 0.99443 0.8666
## Pos Pred Value 0.8455 0.91492 0.9486
## Neg Pred Value 0.9574 0.99560 0.8622
## Prevalence 0.2174 0.06036 0.7222
## Detection Rate 0.1841 0.05623 0.6837
## Detection Prevalence 0.2178 0.06146 0.7208
## Balanced Accuracy 0.9019 0.96299 0.9067
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13941, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9378959 0.8477714
## 2 extratrees 0.8965778 0.7283916
## 7 gini 0.9543579 0.8915981
## 7 extratrees 0.9513878 0.8830350
## 12 gini 0.9524855 0.8872566
## 12 extratrees 0.9535186 0.8889409
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 7, splitrule = gini
## and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2988 1 233
## 1 0 868 27
## 2 380 66 10927
##
## Overall Statistics
##
## Accuracy : 0.9544
## 95% CI : (0.951, 0.9576)
## No Information Rate : 0.7222
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8916
##
## Mcnemar's Test P-Value : 2.224e-11
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8872 0.92834 0.9768
## Specificity 0.9807 0.99814 0.8964
## Pos Pred Value 0.9274 0.96983 0.9608
## Neg Pred Value 0.9690 0.99541 0.9368
## Prevalence 0.2174 0.06036 0.7222
## Detection Rate 0.1929 0.05604 0.7054
## Detection Prevalence 0.2080 0.05778 0.7342
## Balanced Accuracy 0.9339 0.96324 0.9366
## [1] "----------------------------------------------------------------------------"
## [1] "6 13 245"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13941, 13941, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9255622 0.8130173
## 2 0.9045822 0.7612446
## 3 0.9101327 0.7718873
## 4 0.9000643 0.7465836
## 5 0.9005807 0.7455008
## 6 0.8941893 0.7290577
## 7 0.8932216 0.7239793
## 8 0.8897347 0.7143854
## 9 0.8864420 0.7044479
## 10 0.8823753 0.6934154
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2860 1 516
## 1 1 532 66
## 2 507 62 10945
##
## Overall Statistics
##
## Accuracy : 0.9256
## 95% CI : (0.9213, 0.9297)
## No Information Rate : 0.7442
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.813
##
## Mcnemar's Test P-Value : 0.9769
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8492 0.89412 0.9495
## Specificity 0.9574 0.99550 0.8564
## Pos Pred Value 0.8469 0.88815 0.9506
## Neg Pred Value 0.9581 0.99577 0.8536
## Prevalence 0.2174 0.03841 0.7442
## Detection Rate 0.1846 0.03434 0.7066
## Detection Prevalence 0.2180 0.03867 0.7433
## Balanced Accuracy 0.9033 0.94481 0.9030
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13942, 13940, 13940, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9365380 0.8309808
## 2 extratrees 0.8937362 0.6933053
## 7 gini 0.9540357 0.8818446
## 7 extratrees 0.9511296 0.8729086
## 12 gini 0.9497752 0.8711616
## 12 extratrees 0.9553267 0.8845920
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2933 1 191
## 1 0 541 12
## 2 435 53 11324
##
## Overall Statistics
##
## Accuracy : 0.9553
## 95% CI : (0.952, 0.9585)
## No Information Rate : 0.7442
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8847
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8708 0.90924 0.9824
## Specificity 0.9842 0.99919 0.8769
## Pos Pred Value 0.9386 0.97830 0.9587
## Neg Pred Value 0.9648 0.99638 0.9448
## Prevalence 0.2174 0.03841 0.7442
## Detection Rate 0.1893 0.03493 0.7311
## Detection Prevalence 0.2017 0.03570 0.7626
## Balanced Accuracy 0.9275 0.95422 0.9296
## [1] "----------------------------------------------------------------------------"
## [1] "6 14 235"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13941, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8892185 0.8114103
## 2 0.8602317 0.7621287
## 3 0.8659126 0.7705067
## 4 0.8520981 0.7464721
## 5 0.8521634 0.7460455
## 6 0.8421560 0.7280908
## 7 0.8388635 0.7212901
## 8 0.8335054 0.7114521
## 9 0.8298252 0.7047647
## 10 0.8241441 0.6942165
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2868 95 423
## 1 88 2949 350
## 2 412 348 7957
##
## Overall Statistics
##
## Accuracy : 0.8892
## 95% CI : (0.8842, 0.8941)
## No Information Rate : 0.5636
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8114
##
## Mcnemar's Test P-Value : 0.9364
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8515 0.8694 0.9115
## Specificity 0.9573 0.9638 0.8876
## Pos Pred Value 0.8470 0.8707 0.9128
## Neg Pred Value 0.9587 0.9634 0.8859
## Prevalence 0.2174 0.2190 0.5636
## Detection Rate 0.1852 0.1904 0.5137
## Detection Prevalence 0.2186 0.2187 0.5628
## Balanced Accuracy 0.9044 0.9166 0.8995
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13940, 13941, 13942, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9163974 0.8535631
## 2 extratrees 0.8639768 0.7522267
## 7 gini 0.9353140 0.8886410
## 7 extratrees 0.9359587 0.8889844
## 12 gini 0.9276960 0.8757267
## 12 extratrees 0.9386052 0.8939297
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 3018 65 160
## 1 32 3032 81
## 2 318 295 8489
##
## Overall Statistics
##
## Accuracy : 0.9386
## 95% CI : (0.9347, 0.9423)
## No Information Rate : 0.5636
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8939
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8961 0.8939 0.9724
## Specificity 0.9814 0.9907 0.9093
## Pos Pred Value 0.9306 0.9641 0.9327
## Neg Pred Value 0.9714 0.9708 0.9623
## Prevalence 0.2174 0.2190 0.5636
## Detection Rate 0.1948 0.1957 0.5480
## Detection Prevalence 0.2094 0.2030 0.5876
## Balanced Accuracy 0.9388 0.9423 0.9409
## [1] "----------------------------------------------------------------------------"
## [1] "6 15 234"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13942, 13941, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8908977 0.8222845
## 2 0.8595861 0.7714262
## 3 0.8692050 0.7864890
## 4 0.8577139 0.7675636
## 5 0.8556483 0.7637498
## 6 0.8461585 0.7479246
## 7 0.8422199 0.7413335
## 8 0.8367321 0.7318060
## 9 0.8344730 0.7276077
## 10 0.8270477 0.7149629
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2858 433 100
## 1 416 7293 294
## 2 94 353 3649
##
## Overall Statistics
##
## Accuracy : 0.8909
## 95% CI : (0.8859, 0.8958)
## No Information Rate : 0.5216
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8223
##
## Mcnemar's Test P-Value : 0.1163
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8486 0.9027 0.9025
## Specificity 0.9560 0.9042 0.9610
## Pos Pred Value 0.8428 0.9113 0.8909
## Neg Pred Value 0.9578 0.8950 0.9654
## Prevalence 0.2174 0.5216 0.2610
## Detection Rate 0.1845 0.4708 0.2356
## Detection Prevalence 0.2189 0.5167 0.2644
## Balanced Accuracy 0.9023 0.9035 0.9317
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13942, 13942, 13941, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9216278 0.8700753
## 2 extratrees 0.8744368 0.7865860
## 7 gini 0.9401557 0.9018611
## 7 extratrees 0.9381546 0.8980082
## 12 gini 0.9311177 0.8871959
## 12 extratrees 0.9416408 0.9040109
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 3029 168 60
## 1 309 7799 225
## 2 30 112 3758
##
## Overall Statistics
##
## Accuracy : 0.9416
## 95% CI : (0.9378, 0.9453)
## No Information Rate : 0.5216
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.904
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8993 0.9653 0.9295
## Specificity 0.9812 0.9279 0.9876
## Pos Pred Value 0.9300 0.9359 0.9636
## Neg Pred Value 0.9723 0.9609 0.9754
## Prevalence 0.2174 0.5216 0.2610
## Detection Rate 0.1955 0.5035 0.2426
## Detection Prevalence 0.2103 0.5380 0.2518
## Balanced Accuracy 0.9403 0.9466 0.9586
## [1] "----------------------------------------------------------------------------"
## [1] "6 23 145"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13942, 13942, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9200136 0.8138005
## 2 0.8984514 0.7640849
## 3 0.9047771 0.7772318
## 4 0.8930921 0.7490155
## 5 0.8926402 0.7467552
## 6 0.8841822 0.7268668
## 7 0.8827626 0.7214406
## 8 0.8754674 0.7035420
## 9 0.8741116 0.6990309
## 10 0.8717223 0.6928042
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2865 2 519
## 1 3 845 118
## 2 500 97 10541
##
## Overall Statistics
##
## Accuracy : 0.92
## 95% CI : (0.9156, 0.9242)
## No Information Rate : 0.7216
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8139
##
## Mcnemar's Test P-Value : 0.4565
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8507 0.89513 0.9430
## Specificity 0.9570 0.99168 0.8615
## Pos Pred Value 0.8461 0.87474 0.9464
## Neg Pred Value 0.9584 0.99318 0.8536
## Prevalence 0.2174 0.06094 0.7216
## Detection Rate 0.1850 0.05455 0.6805
## Detection Prevalence 0.2186 0.06236 0.7190
## Balanced Accuracy 0.9038 0.94340 0.9023
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13942, 13941, 13942, 13941, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9349893 0.8400683
## 2 extratrees 0.8905739 0.7104764
## 7 gini 0.9512588 0.8838748
## 7 extratrees 0.9500967 0.8797945
## 12 gini 0.9466751 0.8730654
## 12 extratrees 0.9537119 0.8893104
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2944 0 177
## 1 0 864 36
## 2 424 80 10965
##
## Overall Statistics
##
## Accuracy : 0.9537
## 95% CI : (0.9503, 0.957)
## No Information Rate : 0.7216
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8894
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8741 0.91525 0.9809
## Specificity 0.9854 0.99753 0.8831
## Pos Pred Value 0.9433 0.96000 0.9561
## Neg Pred Value 0.9657 0.99452 0.9470
## Prevalence 0.2174 0.06094 0.7216
## Detection Rate 0.1901 0.05578 0.7079
## Detection Prevalence 0.2015 0.05810 0.7404
## Balanced Accuracy 0.9298 0.95639 0.9320
## [1] "----------------------------------------------------------------------------"
## [1] "6 24 135"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13941, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8916735 0.8201001
## 2 0.8643650 0.7753823
## 3 0.8693367 0.7821527
## 4 0.8546171 0.7574907
## 5 0.8562313 0.7598196
## 6 0.8447399 0.7405572
## 7 0.8420278 0.7354187
## 8 0.8378960 0.7278018
## 9 0.8344744 0.7214189
## 10 0.8286636 0.7112662
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2874 87 426
## 1 82 3325 342
## 2 412 329 7613
##
## Overall Statistics
##
## Accuracy : 0.8917
## 95% CI : (0.8867, 0.8965)
## No Information Rate : 0.5411
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8201
##
## Mcnemar's Test P-Value : 0.8887
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8533 0.8888 0.9084
## Specificity 0.9577 0.9639 0.8958
## Pos Pred Value 0.8485 0.8869 0.9113
## Neg Pred Value 0.9592 0.9646 0.8924
## Prevalence 0.2174 0.2415 0.5411
## Detection Rate 0.1855 0.2147 0.4915
## Detection Prevalence 0.2187 0.2420 0.5393
## Balanced Accuracy 0.9055 0.9264 0.9021
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13940, 13941, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9204000 0.8649010
## 2 extratrees 0.8732732 0.7779977
## 7 gini 0.9381536 0.8965162
## 7 extratrees 0.9369912 0.8938166
## 12 gini 0.9291158 0.8814358
## 12 extratrees 0.9428019 0.9039613
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 3048 62 161
## 1 30 3422 86
## 2 290 257 8134
##
## Overall Statistics
##
## Accuracy : 0.9428
## 95% CI : (0.939, 0.9464)
## No Information Rate : 0.5411
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.904
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9050 0.9147 0.9705
## Specificity 0.9816 0.9901 0.9231
## Pos Pred Value 0.9318 0.9672 0.9370
## Neg Pred Value 0.9738 0.9733 0.9637
## Prevalence 0.2174 0.2415 0.5411
## Detection Rate 0.1968 0.2209 0.5251
## Detection Prevalence 0.2112 0.2284 0.5604
## Balanced Accuracy 0.9433 0.9524 0.9468
## [1] "----------------------------------------------------------------------------"
## [1] "6 25 134"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13940, 13941, 13943, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8904461 0.8175082
## 2 0.8595854 0.7662529
## 3 0.8650748 0.7742409
## 4 0.8559729 0.7586003
## 5 0.8546169 0.7562524
## 6 0.8448699 0.7393430
## 7 0.8409960 0.7323714
## 8 0.8336365 0.7193789
## 9 0.8315057 0.7150733
## 10 0.8280186 0.7091497
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2873 424 89
## 1 411 7655 340
## 2 84 349 3265
##
## Overall Statistics
##
## Accuracy : 0.8904
## 95% CI : (0.8854, 0.8953)
## No Information Rate : 0.5441
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8175
##
## Mcnemar's Test P-Value : 0.9266
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8530 0.9083 0.8839
## Specificity 0.9577 0.8937 0.9633
## Pos Pred Value 0.8485 0.9107 0.8829
## Neg Pred Value 0.9591 0.8909 0.9636
## Prevalence 0.2174 0.5441 0.2385
## Detection Rate 0.1855 0.4942 0.2108
## Detection Prevalence 0.2186 0.5427 0.2387
## Balanced Accuracy 0.9054 0.9010 0.9236
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13941, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9163984 0.8574208
## 2 extratrees 0.8692075 0.7701578
## 7 gini 0.9377672 0.8954455
## 7 extratrees 0.9370576 0.8935650
## 12 gini 0.9295037 0.8817537
## 12 extratrees 0.9406725 0.9000565
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 3038 159 65
## 1 303 8165 261
## 2 27 104 3368
##
## Overall Statistics
##
## Accuracy : 0.9407
## 95% CI : (0.9368, 0.9443)
## No Information Rate : 0.5441
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9001
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9020 0.9688 0.9117
## Specificity 0.9815 0.9201 0.9889
## Pos Pred Value 0.9313 0.9354 0.9626
## Neg Pred Value 0.9730 0.9611 0.9728
## Prevalence 0.2174 0.5441 0.2385
## Detection Rate 0.1961 0.5271 0.2174
## Detection Prevalence 0.2106 0.5635 0.2259
## Balanced Accuracy 0.9418 0.9445 0.9503
## [1] "----------------------------------------------------------------------------"
## [1] "6 34 125"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13941, 13942, 13941, 13940, 13940, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8921252 0.8167930
## 2 0.8623640 0.7664080
## 3 0.8712075 0.7801139
## 4 0.8599752 0.7609294
## 5 0.8577157 0.7562837
## 6 0.8466753 0.7371817
## 7 0.8464170 0.7357410
## 8 0.8372499 0.7195863
## 9 0.8358936 0.7165758
## 10 0.8311175 0.7079905
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2876 83 438
## 1 86 3001 341
## 2 406 317 7942
##
## Overall Statistics
##
## Accuracy : 0.8921
## 95% CI : (0.8871, 0.897)
## No Information Rate : 0.563
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8168
##
## Mcnemar's Test P-Value : 0.5435
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8539 0.8824 0.9107
## Specificity 0.9570 0.9647 0.8932
## Pos Pred Value 0.8466 0.8754 0.9166
## Neg Pred Value 0.9593 0.9668 0.8859
## Prevalence 0.2174 0.2196 0.5630
## Detection Rate 0.1857 0.1937 0.5127
## Detection Prevalence 0.2193 0.2213 0.5594
## Balanced Accuracy 0.9055 0.9235 0.9019
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13941, 13942, 13941, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9180769 0.8568112
## 2 extratrees 0.8670756 0.7587016
## 7 gini 0.9363459 0.8906586
## 7 extratrees 0.9378312 0.8924007
## 12 gini 0.9292450 0.8785835
## 12 extratrees 0.9413172 0.8988323
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 3018 62 160
## 1 31 3086 84
## 2 319 253 8477
##
## Overall Statistics
##
## Accuracy : 0.9413
## 95% CI : (0.9375, 0.945)
## No Information Rate : 0.563
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8989
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8961 0.9074 0.9720
## Specificity 0.9817 0.9905 0.9155
## Pos Pred Value 0.9315 0.9641 0.9368
## Neg Pred Value 0.9714 0.9744 0.9621
## Prevalence 0.2174 0.2196 0.5630
## Detection Rate 0.1948 0.1992 0.5473
## Detection Prevalence 0.2092 0.2066 0.5842
## Balanced Accuracy 0.9389 0.9489 0.9438
## [1] "----------------------------------------------------------------------------"
## [1] "6 35 124"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13941, 13940, 13942, 13941, 13942, 13941, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.8916726 0.8232221
## 2 0.8629451 0.7766134
## 3 0.8713363 0.7893048
## 4 0.8591985 0.7694420
## 5 0.8584255 0.7679380
## 6 0.8468706 0.7486481
## 7 0.8466122 0.7476251
## 8 0.8398981 0.7363922
## 9 0.8378961 0.7327547
## 10 0.8328605 0.7241549
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2865 439 98
## 1 414 7328 317
## 2 89 321 3619
##
## Overall Statistics
##
## Accuracy : 0.8917
## 95% CI : (0.8867, 0.8965)
## No Information Rate : 0.5221
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8232
##
## Mcnemar's Test P-Value : 0.7552
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8507 0.9060 0.8971
## Specificity 0.9557 0.9012 0.9642
## Pos Pred Value 0.8422 0.9093 0.8982
## Neg Pred Value 0.9584 0.8977 0.9638
## Prevalence 0.2174 0.5221 0.2604
## Detection Rate 0.1850 0.4731 0.2336
## Detection Prevalence 0.2196 0.5203 0.2601
## Balanced Accuracy 0.9032 0.9036 0.9307
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13942, 13942, 13942, 13940, 13941, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9205939 0.8680473
## 2 extratrees 0.8739816 0.7851272
## 7 gini 0.9391881 0.9001734
## 7 extratrees 0.9385407 0.8986125
## 12 gini 0.9316357 0.8879048
## 12 extratrees 0.9428015 0.9058384
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 3037 165 60
## 1 303 7823 230
## 2 28 100 3744
##
## Overall Statistics
##
## Accuracy : 0.9428
## 95% CI : (0.939, 0.9464)
## No Information Rate : 0.5221
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9058
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.9017 0.9672 0.9281
## Specificity 0.9814 0.9280 0.9888
## Pos Pred Value 0.9310 0.9362 0.9669
## Neg Pred Value 0.9729 0.9629 0.9750
## Prevalence 0.2174 0.5221 0.2604
## Detection Rate 0.1961 0.5050 0.2417
## Detection Prevalence 0.2106 0.5394 0.2500
## Balanced Accuracy 0.9416 0.9476 0.9585
## [1] "----------------------------------------------------------------------------"
## [1] "6 45 123"
## k-Nearest Neighbors
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13942, 13941, 13942, 13943, ...
## Resampling results across tuning parameters:
##
## k Accuracy Kappa
## 1 0.9227890 0.8303628
## 2 0.9027757 0.7863682
## 3 0.9098784 0.7999655
## 4 0.8965793 0.7701836
## 5 0.8987098 0.7736084
## 6 0.8900591 0.7544985
## 7 0.8866374 0.7457181
## 8 0.8805686 0.7316033
## 9 0.8808249 0.7306180
## 10 0.8772104 0.7221207
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2868 536 2
## 1 496 10256 65
## 2 4 93 1170
##
## Overall Statistics
##
## Accuracy : 0.9228
## 95% CI : (0.9185, 0.9269)
## No Information Rate : 0.7027
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.8303
##
## Mcnemar's Test P-Value : 0.0664
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8515 0.9422 0.94584
## Specificity 0.9556 0.8782 0.99319
## Pos Pred Value 0.8420 0.9481 0.92344
## Neg Pred Value 0.9586 0.8654 0.99529
## Prevalence 0.2174 0.7027 0.07986
## Detection Rate 0.1852 0.6621 0.07553
## Detection Prevalence 0.2199 0.6983 0.08179
## Balanced Accuracy 0.9036 0.9102 0.96952
## Random Forest
##
## 15490 samples
## 12 predictor
## 3 classes: '0', '1', '2'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 13940, 13941, 13940, 13942, 13941, 13942, ...
## Resampling results across tuning parameters:
##
## mtry splitrule Accuracy Kappa
## 2 gini 0.9367985 0.8540829
## 2 extratrees 0.8956748 0.7446318
## 7 gini 0.9539698 0.8966729
## 7 extratrees 0.9518400 0.8908731
## 12 gini 0.9497741 0.8873300
## 12 extratrees 0.9548738 0.8983049
##
## Tuning parameter 'min.node.size' was held constant at a value of 1
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were mtry = 12, splitrule =
## extratrees and min.node.size = 1.
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1 2
## 0 2943 191 1
## 1 425 10670 58
## 2 0 24 1178
##
## Overall Statistics
##
## Accuracy : 0.9549
## 95% CI : (0.9515, 0.9581)
## No Information Rate : 0.7027
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.8983
##
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: 0 Class: 1 Class: 2
## Sensitivity 0.8738 0.9802 0.95230
## Specificity 0.9842 0.8951 0.99832
## Pos Pred Value 0.9388 0.9567 0.98003
## Neg Pred Value 0.9656 0.9504 0.99587
## Prevalence 0.2174 0.7027 0.07986
## Detection Rate 0.1900 0.6888 0.07605
## Detection Prevalence 0.2024 0.7200 0.07760
## Balanced Accuracy 0.9290 0.9377 0.97531
## [1] "----------------------------------------------------------------------------"#print(arr_accuracy_knn)
#confusionMatrix(knn)
#print(arr_accuracy_svm)
#confusionMatrix(svm_radial)
#print(arr_accuracy_rf)
#confusionMatrix(rf)
cbind(possible_comb, arr_accuracy_knn, arr_accuracy_rf)## possible_comb arr_accuracy_knn arr_accuracy_rf
## [1,] "1234 5 6" "0.895739186571982" "0.944286636539703"
## [2,] "1235 4 6" "0.890768237572628" "0.93989670755326"
## [3,] "1245 3 6" "0.929502905100065" "0.954938670109748"
## [4,] "1345 2 6" "0.922401549386701" "0.956488056810846"
## [5,] "2345 1 6" "0.928147191736604" "0.95823111684958"
## [6,] "1236 4 5" "0.888508715300194" "0.942091672046482"
## [7,] "1246 3 5" "0.905616526791478" "0.950871530019367"
## [8,] "1346 2 5" "0.897546804389929" "0.949838605551969"
## [9,] "2346 1 5" "0.902517753389284" "0.950806972240155"
## [10,] "1256 3 4" "0.944544867656553" "0.966623628147192"
## [11,] "1356 2 4" "0.941639767591995" "0.966752743705617"
## [12,] "2356 1 4" "0.944544867656553" "0.965590703679793"
## [13,] "1456 2 3" "0.985474499677211" "0.990897353131052"
## [14,] "2456 1 3" "0.990187217559716" "0.993673337637185"
## [15,] "3456 1 2" "0.988315041962556" "0.993673337637185"
## [16,] "1 23 456" "0.9848934796643" "0.991284699806327"
## [17,] "1 24 356" "0.943963847643641" "0.967979341510652"
## [18,] "1 25 346" "0.899870884441575" "0.950484183344093"
## [19,] "1 26 345" "0.921949644932214" "0.955648805681085"
## [20,] "1 34 256" "0.942995480955455" "0.966494512588767"
## [21,] "1 35 246" "0.90135571336346" "0.949903163331181"
## [22,] "1 36 245" "0.923692704970949" "0.955777921239509"
## [23,] "1 45 236" "0.920464816010329" "0.9536475145255"
## [24,] "1 46 235" "0.900581020012912" "0.947579083279535"
## [25,] "1 56 234" "0.945255003227889" "0.968689477081988"
## [26,] "2 13 456" "0.985603615235636" "0.991672046481601"
## [27,] "2 14 356" "0.938928340865074" "0.967269205939316"
## [28,] "2 15 346" "0.894964493221433" "0.949838605551969"
## [29,] "2 16 345" "0.920787604906391" "0.955132343447385"
## [30,] "2 34 156" "0.94151065203357" "0.966881859264041"
## [31,] "2 35 146" "0.898579728857327" "0.950161394448031"
## [32,] "2 36 145" "0.917882504841834" "0.952420916720465"
## [33,] "2 45 136" "0.917753389283409" "0.953195610071013"
## [34,] "2 46 135" "0.899289864428664" "0.94744996772111"
## [35,] "2 56 134" "0.942801807617818" "0.968173014848289"
## [36,] "3 12 456" "0.987475790832795" "0.992640413169787"
## [37,] "3 14 256" "0.942027114267269" "0.966365397030342"
## [38,] "3 15 246" "0.899354422207876" "0.950484183344093"
## [39,] "3 16 245" "0.924144609425436" "0.95551969012266"
## [40,] "3 24 156" "0.944415752098128" "0.967333763718528"
## [41,] "3 25 146" "0.899483537766301" "0.946029696578438"
## [42,] "3 26 145" "0.919690122659781" "0.951710781149129"
## [43,] "3 45 126" "0.922336991607489" "0.954486765655261"
## [44,] "3 46 125" "0.902646868947708" "0.948870238863783"
## [45,] "3 56 124" "0.946417043253712" "0.968560361523564"
## [46,] "4 12 356" "0.941446094254358" "0.966688185926404"
## [47,] "4 13 256" "0.942027114267269" "0.965978050355068"
## [48,] "4 15 236" "0.884828921885087" "0.939122014202711"
## [49,] "4 16 235" "0.888444157520981" "0.940219496449322"
## [50,] "4 23 156" "0.939122014202711" "0.966752743705616"
## [51,] "4 25 136" "0.885409941897999" "0.93989670755326"
## [52,] "4 26 135" "0.887023886378309" "0.940219496449322"
## [53,] "4 35 126" "0.890058102001291" "0.939767591994835"
## [54,] "4 36 125" "0.890187217559716" "0.939315687540349"
## [55,] "4 56 123" "0.943382827630729" "0.969205939315688"
## [56,] "5 12 346" "0.90025823111685" "0.949257585539057"
## [57,] "5 13 246" "0.901743060038735" "0.950871530019367"
## [58,] "5 14 236" "0.885474499677211" "0.939638476436411"
## [59,] "5 16 234" "0.891220142027114" "0.941962556488057"
## [60,] "5 23 146" "0.897675919948354" "0.948999354422208"
## [61,] "5 24 136" "0.887475790832795" "0.942091672046482"
## [62,] "5 26 134" "0.888896061975468" "0.939832149774048"
## [63,] "5 34 126" "0.889606197546804" "0.941058747579083"
## [64,] "5 36 124" "0.891930277598451" "0.94125242091672"
## [65,] "5 46 123" "0.899225306649451" "0.950290510006456"
## [66,] "6 12 345" "0.924080051646223" "0.954357650096837"
## [67,] "6 13 245" "0.925564880568108" "0.955326016785023"
## [68,] "6 14 235" "0.88921885087153" "0.938605551969012"
## [69,] "6 15 234" "0.890897353131052" "0.941639767591995"
## [70,] "6 23 145" "0.920012911555842" "0.953712072304713"
## [71,] "6 24 135" "0.891672046481601" "0.942801807617818"
## [72,] "6 25 134" "0.890445448676565" "0.940671400903809"
## [73,] "6 34 125" "0.892123950936088" "0.941316978695933"
## [74,] "6 35 124" "0.891672046481601" "0.942801807617818"
## [75,] "6 45 123" "0.922788896061976" "0.954874112330536"