Supervised Machine Learning

1. KNN - WINE

Caret package

trainControl() - 훈련과정 중 parameter 설정 Ex) trainControl ( method = “repeatedcv” number = 10, <- 훈련데이터 fold 갯수 repeat - 5) <- cv 반복 횟수

expand.grid() - factor 조합의 데이터 프레임 생성 Ex) expand.grid (k=1:10)

train() Ex) train ( Class ~. , data = method= trContraol preProcess = c(“center”,“scale”), <- 표준화, tuneGrid = expand.grid(k=1:10), <- 파라미터값 목록 metric="Accuarcy) <- 모형방식

Accurarcy 정확도 = TP+TN/ Total

Kappa 통계량 = p0-pe/1-pe p0: 관측 정확도 / pe: 기대 정확도

setwd('C:/Users/Administrator/Desktop/R Analysis/Fast Campus/Part 6/Ch02. k-Nearest Neighbor/Data')


read.csv("wine.csv",header = TRUE) -> rawdata

rawdata$Class <- as.factor(rawdata$Class) #factor 변환 
str(rawdata)

## 'data.frame':    178 obs. of  14 variables:
##  $ Alcohol             : num  14.2 13.2 13.2 14.4 13.2 ...
##  $ Acid                : num  1.71 1.78 2.36 1.95 2.59 1.76 1.87 2.15 1.64 1.35 ...
##  $ Ash                 : num  2.43 2.14 2.67 2.5 2.87 2.45 2.45 2.61 2.17 2.27 ...
##  $ Alcalinity          : num  15.6 11.2 18.6 16.8 21 15.2 14.6 17.6 14 16 ...
##  $ Magnesium           : int  127 100 101 113 118 112 96 121 97 98 ...
##  $ Total_phenols       : num  2.8 2.65 2.8 3.85 2.8 3.27 2.5 2.6 2.8 2.98 ...
##  $ Flavanoids          : num  3.06 2.76 3.24 3.49 2.69 3.39 2.52 2.51 2.98 3.15 ...
##  $ Nonflavanoid_phenols: num  0.28 0.26 0.3 0.24 0.39 0.34 0.3 0.31 0.29 0.22 ...
##  $ Proanthocyanins     : num  2.29 1.28 2.81 2.18 1.82 1.97 1.98 1.25 1.98 1.85 ...
##  $ color_intensity     : num  5.64 4.38 5.68 7.8 4.32 6.75 5.25 5.05 5.2 7.22 ...
##  $ Hue                 : num  1.04 1.05 1.03 0.86 1.04 1.05 1.02 1.06 1.08 1.01 ...
##  $ X0D280              : num  3.92 3.4 3.17 3.45 2.93 2.85 3.58 3.58 2.85 3.55 ...
##  $ proline             : int  1065 1050 1185 1480 735 1450 1290 1295 1045 1045 ...
##  $ Class               : Factor w/ 3 levels "1","2","3": 1 1 1 1 1 1 1 1 1 1 ...

datatable(rawdata)

Train, Test 분할

analdata <-rawdata


set.seed(2020) #Seed 설정 
datatotal <- sort(sample(nrow(analdata), nrow(analdata)*0.7)) #sample 표본 뽑기 100중에 70개 뽑기  

#sample(a,b) : 1부터 a까지 숫자 중에 b개 추출 

train <-rawdata[datatotal,]
test <-rawdata[-datatotal, ]


train_x <-train[,1:13]
train_y <-train[,14]

test_x <-test[,1:13]
test_y <-test[,14]

모형

ctrl <-trainControl(method ="repeatedcv", 
                    number =10, 
                    repeats=5)


customGrid <-expand.grid(k=1:10)

train(Class~., 
      data=train,
      method = "knn", 
      trControl = ctrl, 
      preProcess= c("center","scale"), 
      tuneGrid = customGrid, 
      metric="Accuracy") -> knnFit

knnFit

## k-Nearest Neighbors 
## 
## 124 samples
##  13 predictor
##   3 classes: '1', '2', '3' 
## 
## Pre-processing: centered (13), scaled (13) 
## Resampling: Cross-Validated (10 fold, repeated 5 times) 
## Summary of sample sizes: 111, 112, 112, 111, 110, 112, ... 
## Resampling results across tuning parameters:
## 
##   k   Accuracy   Kappa    
##    1  0.9726041  0.9585595
##    2  0.9658725  0.9482692
##    3  0.9707859  0.9558296
##    4  0.9681901  0.9520425
##    5  0.9776956  0.9662325
##    6  0.9603280  0.9401272
##    7  0.9567100  0.9348356
##    8  0.9517965  0.9274056
##    9  0.9644023  0.9465077
##   10  0.9673876  0.9507825
## 
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was k = 5.

plot(knnFit)

Prediction

pred_test <-predict(knnFit, newdata = test)

confusionMatrix(pred_test, test$Class)

## Confusion Matrix and Statistics
## 
##           Reference
## Prediction  1  2  3
##          1 14  2  0
##          2  0 24  0
##          3  0  1 13
## 
## Overall Statistics
##                                           
##                Accuracy : 0.9444          
##                  95% CI : (0.8461, 0.9884)
##     No Information Rate : 0.5             
##     P-Value [Acc > NIR] : 1.459e-12       
##                                           
##                   Kappa : 0.913           
##                                           
##  Mcnemar's Test P-Value : NA              
## 
## Statistics by Class:
## 
##                      Class: 1 Class: 2 Class: 3
## Sensitivity            1.0000   0.8889   1.0000
## Specificity            0.9500   1.0000   0.9756
## Pos Pred Value         0.8750   1.0000   0.9286
## Neg Pred Value         1.0000   0.9000   1.0000
## Prevalence             0.2593   0.5000   0.2407
## Detection Rate         0.2593   0.4444   0.2407
## Detection Prevalence   0.2963   0.4444   0.2593
## Balanced Accuracy      0.9750   0.9444   0.9878

Importance Feature

varImp(knnFit, scale = FALSE) ->importance_knn
plot(importance_knn)

2. Logistic - HeartAttack

setwd('C:/Users/Administrator/Desktop/R Analysis/Fast Campus/Part 6/Ch03. Logistic Regression/Data')
read.csv("heart.csv", header = TRUE) -> heart

str(heart)

## 'data.frame':    303 obs. of  14 variables:
##  $ age     : int  63 37 41 56 57 57 56 44 52 57 ...
##  $ sex     : int  1 1 0 1 0 1 0 1 1 1 ...
##  $ cp      : int  3 2 1 1 0 0 1 1 2 2 ...
##  $ trestbps: int  145 130 130 120 120 140 140 120 172 150 ...
##  $ chol    : int  233 250 204 236 354 192 294 263 199 168 ...
##  $ fbs     : int  1 0 0 0 0 0 0 0 1 0 ...
##  $ restecg : int  0 1 0 1 1 1 0 1 1 1 ...
##  $ thalach : int  150 187 172 178 163 148 153 173 162 174 ...
##  $ exang   : int  0 0 0 0 1 0 0 0 0 0 ...
##  $ oldpeak : num  2.3 3.5 1.4 0.8 0.6 0.4 1.3 0 0.5 1.6 ...
##  $ slope   : int  0 0 2 2 2 1 1 2 2 2 ...
##  $ ca      : int  0 0 0 0 0 0 0 0 0 0 ...
##  $ thal    : int  1 2 2 2 2 1 2 3 3 2 ...
##  $ target  : int  1 1 1 1 1 1 1 1 1 1 ...

#0과 1로 나타난 변수들의 전처리가 필요함 

heart$target <-as.factor(heart$target)
unique(heart$target)

## [1] 1 0
## Levels: 0 1

#---------------------------------------------------------
# 연속형 변수는 표준화 
#--------------------------------------------------------

heart$age   <-   scale(heart$age)
heart$trestbps  <-  scale(heart$trestbps)
heart$chol<-  scale(heart$chol)
heart$thalach <- scale(heart$thalach)
heart$oldpeak<-  scale(heart$oldpeak)
heart$slope <- scale(heart$slope)


#---------------------------------------------------------
# 범주 변수는 as.factor 
#--------------------------------------------------------

newdata<-heart
factorVar <- c("sex", "cp", "restecg", "exang","ca","thal")
newdata[,factorVar] = lapply(newdata[,factorVar], factor)

Tran, Test 나누기

set.seed(2020)

sample <- sort(sample(nrow(newdata),nrow(newdata)*0.7))

train <- newdata[sample,]
test <- newdata[-sample,]

train_x <-train[,1:12]
train_y <- train[,13]

test_x <- test[,1:12]
test_y <- test[,13]

Modelling

ctrl <- trainControl(method = "repeatedcv", repeats = 5)

train(target~., 
      data=train, 
      method = "LogitBoost", 
      trContrl = ctrl, 
      metric = "Accuracy") -> logitFit


logitFit

## Boosted Logistic Regression 
## 
## 212 samples
##  13 predictor
##   2 classes: '0', '1' 
## 
## No pre-processing
## Resampling: Bootstrapped (25 reps) 
## Summary of sample sizes: 212, 212, 212, 212, 212, 212, ... 
## Resampling results across tuning parameters:
## 
##   nIter  Accuracy   Kappa    
##   11     0.8065542  0.5981106
##   21     0.8094922  0.6078944
##   31     0.7919591  0.5711976
## 
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was nIter = 21.

plot(logitFit)

predict(logitFit, newdata = test) ->pred_test

confusionMatrix(pred_test, test$target)

## Confusion Matrix and Statistics
## 
##           Reference
## Prediction  0  1
##          0 33  7
##          1 15 36
##                                           
##                Accuracy : 0.7582          
##                  95% CI : (0.6572, 0.8419)
##     No Information Rate : 0.5275          
##     P-Value [Acc > NIR] : 4.977e-06       
##                                           
##                   Kappa : 0.5197          
##                                           
##  Mcnemar's Test P-Value : 0.1356          
##                                           
##             Sensitivity : 0.6875          
##             Specificity : 0.8372          
##          Pos Pred Value : 0.8250          
##          Neg Pred Value : 0.7059          
##              Prevalence : 0.5275          
##          Detection Rate : 0.3626          
##    Detection Prevalence : 0.4396          
##       Balanced Accuracy : 0.7624          
##                                           
##        'Positive' Class : 0               
## 

Importance Variables

varImp(logitFit, scale = FALSE) -> importance_logit

plot(importance_logit)

3. Naive_bayes - WINE

setwd('C:/Users/Administrator/Desktop/R Analysis/Fast Campus/Part 6/Ch02. k-Nearest Neighbor/Data')


read.csv("wine.csv", header = T) -> wine


wine$Class <- as.factor(wine$Class)
str(wine)

## 'data.frame':    178 obs. of  14 variables:
##  $ Alcohol             : num  14.2 13.2 13.2 14.4 13.2 ...
##  $ Acid                : num  1.71 1.78 2.36 1.95 2.59 1.76 1.87 2.15 1.64 1.35 ...
##  $ Ash                 : num  2.43 2.14 2.67 2.5 2.87 2.45 2.45 2.61 2.17 2.27 ...
##  $ Alcalinity          : num  15.6 11.2 18.6 16.8 21 15.2 14.6 17.6 14 16 ...
##  $ Magnesium           : int  127 100 101 113 118 112 96 121 97 98 ...
##  $ Total_phenols       : num  2.8 2.65 2.8 3.85 2.8 3.27 2.5 2.6 2.8 2.98 ...
##  $ Flavanoids          : num  3.06 2.76 3.24 3.49 2.69 3.39 2.52 2.51 2.98 3.15 ...
##  $ Nonflavanoid_phenols: num  0.28 0.26 0.3 0.24 0.39 0.34 0.3 0.31 0.29 0.22 ...
##  $ Proanthocyanins     : num  2.29 1.28 2.81 2.18 1.82 1.97 1.98 1.25 1.98 1.85 ...
##  $ color_intensity     : num  5.64 4.38 5.68 7.8 4.32 6.75 5.25 5.05 5.2 7.22 ...
##  $ Hue                 : num  1.04 1.05 1.03 0.86 1.04 1.05 1.02 1.06 1.08 1.01 ...
##  $ X0D280              : num  3.92 3.4 3.17 3.45 2.93 2.85 3.58 3.58 2.85 3.55 ...
##  $ proline             : int  1065 1050 1185 1480 735 1450 1290 1295 1045 1045 ...
##  $ Class               : Factor w/ 3 levels "1","2","3": 1 1 1 1 1 1 1 1 1 1 ...

TRAIN/TEST 나누기

samdata <-wine

set.seed(2020)

sort(sample(nrow(samdata), nrow(samdata)*.7))-> samples

train <-wine[samples,]
test <-wine[-samples,]

train_x <-train[,1:13]
train_y <-train[, 14]

test_x <-test[, 1:13]
test_y <-test[,14]

Modelling

ctrl <- trainControl(method="repeatedcv",repeats = 5)
nbFit <- train(Class ~ .,
                 data = train,
                 method = "naive_bayes",
                trControl = ctrl,
                 preProcess = c("center","scale"),
                metric="Accuracy")
nbFit

## Naive Bayes 
## 
## 124 samples
##  13 predictor
##   3 classes: '1', '2', '3' 
## 
## Pre-processing: centered (13), scaled (13) 
## Resampling: Cross-Validated (10 fold, repeated 5 times) 
## Summary of sample sizes: 111, 112, 112, 111, 110, 112, ... 
## Resampling results across tuning parameters:
## 
##   usekernel  Accuracy   Kappa    
##   FALSE      0.9808142  0.9708775
##    TRUE      0.9756810  0.9631746
## 
## Tuning parameter 'laplace' was held constant at a value of 0
## Tuning
##  parameter 'adjust' 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 laplace = 0, usekernel = FALSE
##  and adjust = 1.

plot(nbFit)

커널은 사용 x, 커널 o 에 따른 정확도를 보여줌

predict(nbFit, newdata = test) -> pred_test

confusionMatrix(pred_test, test$Class)

## Confusion Matrix and Statistics
## 
##           Reference
## Prediction  1  2  3
##          1 14  1  0
##          2  0 24  0
##          3  0  2 13
## 
## Overall Statistics
##                                           
##                Accuracy : 0.9444          
##                  95% CI : (0.8461, 0.9884)
##     No Information Rate : 0.5             
##     P-Value [Acc > NIR] : 1.459e-12       
##                                           
##                   Kappa : 0.913           
##                                           
##  Mcnemar's Test P-Value : NA              
## 
## Statistics by Class:
## 
##                      Class: 1 Class: 2 Class: 3
## Sensitivity            1.0000   0.8889   1.0000
## Specificity            0.9750   1.0000   0.9512
## Pos Pred Value         0.9333   1.0000   0.8667
## Neg Pred Value         1.0000   0.9000   1.0000
## Prevalence             0.2593   0.5000   0.2407
## Detection Rate         0.2593   0.4444   0.2407
## Detection Prevalence   0.2778   0.4444   0.2778
## Balanced Accuracy      0.9875   0.9444   0.9756

imp_var <- varImp(nbFit, scale=F)
plot(imp_var)

ROC 커브 기준으로 변수 중요도를 선별해서 보여준다. ROC 커브 면적이 넓을 수록 중요도가 상승한다.

4. Descision Tree

기본 트리 설정

#install.packages("tree")
library(tree)
setwd('C:/Users/Administrator/Desktop/R Analysis/Fast Campus/Part 6/Ch02. k-Nearest Neighbor/Data')


read.csv("wine.csv",header = TRUE) -> rawdata

rawdata$Class <- as.factor(rawdata$Class)

dedata <- rawdata

set.seed(2020)

sample_data <- sort(sample(nrow(dedata), nrow(dedata)*.7))


train <-dedata[sample_data,]
test <-dedata[-sample_data, ]


train_x <-train[,1:13]
train_y <-train[,14]

test_x <-test[,1:13]
test_y <-test[,14]

tree(Class~., data=train) ->treeRaw

plot(treeRaw)
text(treeRaw)

Cross-Validation

cv_tree<-cv.tree(treeRaw, FUN=prune.misclass) #오분류 기준 

plot(cv_tree)

Pruning 가지치기

prune.misclass(treeRaw, best=4) -> prune_tree
plot(prune_tree)
text(prune_tree, pretty=0) #pretty=0 분할 feature 의 이름을 바꾸지 않는다. 

Prediciton

predict(prune_tree, test, type='class') -> pred_test
confusionMatrix(pred_test, test$Class)

## Confusion Matrix and Statistics
## 
##           Reference
## Prediction  1  2  3
##          1 14  2  4
##          2  0 23  0
##          3  0  2  9
## 
## Overall Statistics
##                                           
##                Accuracy : 0.8519          
##                  95% CI : (0.7288, 0.9338)
##     No Information Rate : 0.5             
##     P-Value [Acc > NIR] : 6.922e-08       
##                                           
##                   Kappa : 0.7692          
##                                           
##  Mcnemar's Test P-Value : 0.04601         
## 
## Statistics by Class:
## 
##                      Class: 1 Class: 2 Class: 3
## Sensitivity            1.0000   0.8519   0.6923
## Specificity            0.8500   1.0000   0.9512
## Pos Pred Value         0.7000   1.0000   0.8182
## Neg Pred Value         1.0000   0.8710   0.9070
## Prevalence             0.2593   0.5000   0.2407
## Detection Rate         0.2593   0.4259   0.1667
## Detection Prevalence   0.3704   0.4259   0.2037
## Balanced Accuracy      0.9250   0.9259   0.8218

5. Random Forest

ctrl <- trainControl(method = "repeatedcv", repeats = 5)

train(Class~., 
      data=train, 
      method ="rf", 
      trControl=ctrl, 
      preProcess= c("center", "scale"), 
      metric= "Accuracy") -> rfFit


rfFit

## Random Forest 
## 
## 124 samples
##  13 predictor
##   3 classes: '1', '2', '3' 
## 
## Pre-processing: centered (13), scaled (13) 
## Resampling: Cross-Validated (10 fold, repeated 5 times) 
## Summary of sample sizes: 111, 110, 111, 112, 112, 111, ... 
## Resampling results across tuning parameters:
## 
##   mtry  Accuracy   Kappa    
##    2    0.9904529  0.9855947
##    7    0.9807326  0.9708665
##   13    0.9594539  0.9385166
## 
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was mtry = 2.

varImp(rfFit) -> imp
plot(imp)

6. SVM Linear

setwd('C:/Users/Administrator/Desktop/R Analysis/Fast Campus/Part 6/Ch02. k-Nearest Neighbor/Data')

read.csv("wine.csv",header = TRUE) -> wine

wine$Class <-as.factor(wine$Class)

str(wine)

## 'data.frame':    178 obs. of  14 variables:
##  $ Alcohol             : num  14.2 13.2 13.2 14.4 13.2 ...
##  $ Acid                : num  1.71 1.78 2.36 1.95 2.59 1.76 1.87 2.15 1.64 1.35 ...
##  $ Ash                 : num  2.43 2.14 2.67 2.5 2.87 2.45 2.45 2.61 2.17 2.27 ...
##  $ Alcalinity          : num  15.6 11.2 18.6 16.8 21 15.2 14.6 17.6 14 16 ...
##  $ Magnesium           : int  127 100 101 113 118 112 96 121 97 98 ...
##  $ Total_phenols       : num  2.8 2.65 2.8 3.85 2.8 3.27 2.5 2.6 2.8 2.98 ...
##  $ Flavanoids          : num  3.06 2.76 3.24 3.49 2.69 3.39 2.52 2.51 2.98 3.15 ...
##  $ Nonflavanoid_phenols: num  0.28 0.26 0.3 0.24 0.39 0.34 0.3 0.31 0.29 0.22 ...
##  $ Proanthocyanins     : num  2.29 1.28 2.81 2.18 1.82 1.97 1.98 1.25 1.98 1.85 ...
##  $ color_intensity     : num  5.64 4.38 5.68 7.8 4.32 6.75 5.25 5.05 5.2 7.22 ...
##  $ Hue                 : num  1.04 1.05 1.03 0.86 1.04 1.05 1.02 1.06 1.08 1.01 ...
##  $ X0D280              : num  3.92 3.4 3.17 3.45 2.93 2.85 3.58 3.58 2.85 3.55 ...
##  $ proline             : int  1065 1050 1185 1480 735 1450 1290 1295 1045 1045 ...
##  $ Class               : Factor w/ 3 levels "1","2","3": 1 1 1 1 1 1 1 1 1 1 ...

wine -> svm_wine


set.seed(2021)

wine_sam <- sort(sample(nrow(svm_wine), nrow(svm_wine)*.7))


train <- svm_wine[wine_sam,]
test <- svm_wine[-wine_sam,]



ctrl <- trainControl(method ="repeatedcv", repeats=5)

train(Class~., 
      data=train, 
      method = "svmLinear", #svmPoly 비선형 svm 
      trControl=ctrl, 
      preProcess= c("center","scale"), 
      metric ="Accuracy") ->svm_Fit



svm_Fit

## Support Vector Machines with Linear Kernel 
## 
## 124 samples
##  13 predictor
##   3 classes: '1', '2', '3' 
## 
## Pre-processing: centered (13), scaled (13) 
## Resampling: Cross-Validated (10 fold, repeated 5 times) 
## Summary of sample sizes: 111, 111, 112, 113, 111, 112, ... 
## Resampling results:
## 
##   Accuracy   Kappa    
##   0.9474242  0.9207634
## 
## Tuning parameter 'C' was held constant at a value of 1

predict(svm_Fit, newdata = test) -> pred_test
confusionMatrix(pred_test, test$Class)

## Confusion Matrix and Statistics
## 
##           Reference
## Prediction  1  2  3
##          1 23  0  0
##          2  0 20  0
##          3  0  1 10
## 
## Overall Statistics
##                                           
##                Accuracy : 0.9815          
##                  95% CI : (0.9011, 0.9995)
##     No Information Rate : 0.4259          
##     P-Value [Acc > NIR] : < 2.2e-16       
##                                           
##                   Kappa : 0.9709          
##                                           
##  Mcnemar's Test P-Value : NA              
## 
## Statistics by Class:
## 
##                      Class: 1 Class: 2 Class: 3
## Sensitivity            1.0000   0.9524   1.0000
## Specificity            1.0000   1.0000   0.9773
## Pos Pred Value         1.0000   1.0000   0.9091
## Neg Pred Value         1.0000   0.9706   1.0000
## Prevalence             0.4259   0.3889   0.1852
## Detection Rate         0.4259   0.3704   0.1852
## Detection Prevalence   0.4259   0.3704   0.2037
## Balanced Accuracy      1.0000   0.9762   0.9886

varImp(svm_Fit, scale = FALSE)-> imp
plot(imp)

7. KNN & Logistic

Heat Data 를 KNN 과 Logistic 를 사용해보자

범주형 변수는 factor 형으로 변환 시켜, df_1 테이블 생성

setwd('C:/Users/Administrator/Desktop/R Analysis/Fast Campus/Part 6/Ch03. Logistic Regression/Data')
read.csv("heart.csv", header = TRUE) -> heart

heart %>% 
  mutate(sex= as.factor(sex), 
         cp = as.factor(cp), 
         fbs = as.factor(fbs), 
         exang= as.factor(exang), 
         thal = as.factor(thal), 
         target= as.factor(target)) -> df_1
  

# 결측치 확인해보기 
colSums(is.na(df_1))

##      age      sex       cp trestbps     chol      fbs  restecg  thalach 
##        0        0        0        0        0        0        0        0 
##    exang  oldpeak    slope       ca     thal   target 
##        0        0        0        0        0        0

summary(df_1)

##       age        sex     cp         trestbps          chol       fbs    
##  Min.   :29.00   0: 96   0:143   Min.   : 94.0   Min.   :126.0   0:258  
##  1st Qu.:47.50   1:207   1: 50   1st Qu.:120.0   1st Qu.:211.0   1: 45  
##  Median :55.00           2: 87   Median :130.0   Median :240.0          
##  Mean   :54.37           3: 23   Mean   :131.6   Mean   :246.3          
##  3rd Qu.:61.00                   3rd Qu.:140.0   3rd Qu.:274.5          
##  Max.   :77.00                   Max.   :200.0   Max.   :564.0          
##     restecg          thalach      exang      oldpeak         slope      
##  Min.   :0.0000   Min.   : 71.0   0:204   Min.   :0.00   Min.   :0.000  
##  1st Qu.:0.0000   1st Qu.:133.5   1: 99   1st Qu.:0.00   1st Qu.:1.000  
##  Median :1.0000   Median :153.0           Median :0.80   Median :1.000  
##  Mean   :0.5281   Mean   :149.6           Mean   :1.04   Mean   :1.399  
##  3rd Qu.:1.0000   3rd Qu.:166.0           3rd Qu.:1.60   3rd Qu.:2.000  
##  Max.   :2.0000   Max.   :202.0           Max.   :6.20   Max.   :2.000  
##        ca         thal    target 
##  Min.   :0.0000   0:  2   0:138  
##  1st Qu.:0.0000   1: 18   1:165  
##  Median :0.0000   2:166          
##  Mean   :0.7294   3:117          
##  3rd Qu.:1.0000                  
##  Max.   :4.0000

thal 변수에 0 이 2개 있는데 확인해보기

df_1 %>%  
  filter(thal ==0)

##   age sex cp trestbps chol fbs restecg thalach exang oldpeak slope ca thal
## 1  53   0  2      128  216   0       0     115     0       0     2  0    0
## 2  52   1  0      128  204   1       1     156     1       1     1  0    0
##   target
## 1      1
## 2      0

thal의 0 은 결측 인것으로 판단됨 thal 이 0 이 아닌것만 추출

df_1 %>% 
  filter(thal !=0) -> df_1

###KNN Modelling

KNN 모델에 사용되는 변수들은 숫자형이기 때문에

1. as.numeric 변환
2. scaling 적용

KNN 활용의 특징

1. Heart 데이터는 1,2 로 구분되어 있는 2가지 범주 예측 기존 train 함수를 사용해서는 2 class 분할이 되지 않는다 그래서 KNN 함수를 사용해서 모델링 완료

1. 데이터 분할

2. 분할된 데이터를 숫자형으로 변환

3. 1-13 목표 변수를 제외한 모든 변수를 train/test 마다 scale 를 직접 진행 scale 함수를 사용

4. Target 변수를 labeling 따로 지정(train, test 각각)

5. Optimal K 지정 가능 round(sqrt(nrow ))

6. knn 함수 예측 시작 (train/ test/ cl/ k) 지정

7. Confusion Matrix 정확도 확인

   KNN 함수를 따로 적용하면 따로 따로 해야할 일이 많음 Caret 이 train 함수를 적용할 때에는 3가지 범주는 자동적으로 적용 아니면 이렇게 전부 손수 코드 작성해야한다.

#-----------------------------------------
#   분할 
#------------------------------------------


flag <- sort(sample(nrow(df_1), nrow(df_1)*0.7))

train_h <- df_1[flag,]
test_h <- df_1[-flag,]





#-----------------------------------------
#   분할된  각각의 train, test 에 as.numeric 적용 
#------------------------------------------


train_heart<- train_h %>% 
  mutate(sex = as.numeric(sex),
         cp = as.numeric(cp),
         fbs = as.numeric(fbs),
         exang = as.numeric(exang),
         thal = as.numeric(thal),
         target = as.numeric(target))


test_heart<- test_h %>% 
  mutate(sex = as.numeric(sex),
         cp = as.numeric(cp),
         fbs = as.numeric(fbs),
         exang = as.numeric(exang),
         thal = as.numeric(thal),
         target = as.numeric(target))

#-----------------------------------------
#   Target 변수를 제외한 나머지에 scale 적용  
#------------------------------------------


train <- scale(x=train_heart[, -14])
test <- scale(x=test_heart[,-14], 
              center = attr(train, "scaled:center"), 
              scale = attr(train, "scaled:scale"))


#-----------------------------------------
#   Label 설정 
#------------------------------------------

tr_label <- train_heart[,14]
te_label <- test_heart[,14]


#-----------------------------------------
#   Optimal K = 15 
#------------------------------------------


round(sqrt(nrow(train_heart)),0)

## [1] 14

#-----------------------------------------
#  KNN 함수 사용 
#------------------------------------------


knn(train=train,
    test=test, 
    cl = tr_label,
    k=15) ->KNN_FIT

head(KNN_FIT)

## [1] 2 2 2 2 2 2
## Levels: 1 2

te_label <-as.factor(te_label)


confusionMatrix(data=KNN_FIT, 
                reference = te_label, 
                positive= "2")

## Confusion Matrix and Statistics
## 
##           Reference
## Prediction  1  2
##          1 35  5
##          2 10 41
##                                           
##                Accuracy : 0.8352          
##                  95% CI : (0.7427, 0.9047)
##     No Information Rate : 0.5055          
##     P-Value [Acc > NIR] : 5.352e-11       
##                                           
##                   Kappa : 0.6699          
##                                           
##  Mcnemar's Test P-Value : 0.3017          
##                                           
##             Sensitivity : 0.8913          
##             Specificity : 0.7778          
##          Pos Pred Value : 0.8039          
##          Neg Pred Value : 0.8750          
##              Prevalence : 0.5055          
##          Detection Rate : 0.4505          
##    Detection Prevalence : 0.5604          
##       Balanced Accuracy : 0.8345          
##                                           
##        'Positive' Class : 2               
## 

의견) KNN 은 2 개의 범주형 모델에는 적합한 모델로 보여지지 않는다.