This is an R Markdown document. Markdown is a simple formatting syntax for authoring HTML, PDF, and MS Word documents. For more details on using R Markdown see http://rmarkdown.rstudio.com.
When you click the Knit button a document will be generated that includes both content as well as the output of any embedded R code chunks within the document. You can embed an R code chunk like this:
summary(cars)
## speed dist
## Min. : 4.0 Min. : 2.00
## 1st Qu.:12.0 1st Qu.: 26.00
## Median :15.0 Median : 36.00
## Mean :15.4 Mean : 42.98
## 3rd Qu.:19.0 3rd Qu.: 56.00
## Max. :25.0 Max. :120.00
You can also embed plots, for example:
Note that the echo = FALSE parameter was added to the
code chunk to prevent printing of the R code that generated the
plot.
rm(list = ls())
set.seed(123)
library("gbm")
## Warning: package 'gbm' was built under R version 4.5.3
## Loaded gbm 2.3.1
## This version of gbm is no longer under development. Consider transitioning to gbm3, https://github.com/gbm-developers/gbm3
library("caret")
## Warning: package 'caret' was built under R version 4.5.3
## Loading required package: ggplot2
## Warning: package 'ggplot2' was built under R version 4.5.3
## Loading required package: lattice
library("ggplot2")
# DATASET
data_train <- read.csv("C:/Users/HP/OneDrive/Documents/MUTIA SEMES-5/MACHINE LEARNING/PRAKTIKUM ADABOOST U GRADIENTBOOST/train.csv")
data_test <- read.csv("C:/Users/HP/OneDrive/Documents/MUTIA SEMES-5/MACHINE LEARNING/PRAKTIKUM ADABOOST U GRADIENTBOOST/test.csv")
# PRE-PROCESSING DATA
# Memilih variabel
df <- data_train[, c("price",
"room_type", "city", "accommodates",
"bedrooms", "bathrooms", "number_of_reviews")]
data_test <- data_test[, c("room_type", "city", "accommodates",
"bedrooms", "bathrooms", "number_of_reviews")]
# Memeriksa nilai hilang
colSums(is.na(df))
## price room_type city accommodates
## 0 0 0 0
## bedrooms bathrooms number_of_reviews
## 66 144 0
colSums(is.na(data_test))
## room_type city accommodates bedrooms
## 0 0 0 25
## bathrooms number_of_reviews
## 56 0
# Imputasi nilai hilang dengan median (data train)
med_bedrooms <- median(df$bedrooms, na.rm = TRUE)
med_bathrooms <- median(df$bathrooms, na.rm = TRUE)
med_bedrooms
## [1] 1
med_bathrooms
## [1] 1
df$bedrooms[is.na(df$bedrooms)] <- med_bedrooms
df$bathrooms[is.na(df$bathrooms)] <- med_bathrooms
data_test$bedrooms[is.na(data_test$bedrooms)] <- med_bedrooms
data_test$bathrooms[is.na(data_test$bathrooms)] <- med_bathrooms
# Memeriksa nilai hilang setelah imputasi
colSums(is.na(df))
## price room_type city accommodates
## 0 0 0 0
## bedrooms bathrooms number_of_reviews
## 0 0 0
colSums(is.na(data_test))
## room_type city accommodates bedrooms
## 0 0 0 0
## bathrooms number_of_reviews
## 0 0
# Mengubah variabel kategorik menjadi factor (data train)
df$room_type <- as.factor(df$room_type)
df$city <- as.factor(df$city)
# Mengubah variabel kategorik menjadi factor (data test)
data_test$room_type <- as.factor(data_test$room_type)
data_test$city <- as.factor(data_test$city)
# Memeriksa hasil pre-processing
str(df)
## 'data.frame': 51879 obs. of 7 variables:
## $ price : num 150 169 145 750 115 ...
## $ room_type : Factor w/ 3 levels "Entire home/apt",..: 1 1 1 1 1 2 1 1 2 2 ...
## $ city : Factor w/ 6 levels "Boston","Chicago",..: 5 5 5 6 3 6 4 4 6 4 ...
## $ accommodates : int 3 7 5 4 2 2 3 2 2 2 ...
## $ bedrooms : int 1 3 1 2 0 1 1 1 1 1 ...
## $ bathrooms : num 1 1 1 1 1 1 1 1 1 1 ...
## $ number_of_reviews: int 2 6 10 0 4 3 15 9 159 2 ...
str(data_test)
## 'data.frame': 22232 obs. of 6 variables:
## $ room_type : Factor w/ 3 levels "Entire home/apt",..: 1 2 1 2 1 1 1 2 1 2 ...
## $ city : Factor w/ 6 levels "Boston","Chicago",..: 4 2 4 4 3 5 4 4 5 5 ...
## $ accommodates : int 4 2 4 2 4 6 4 8 2 1 ...
## $ bedrooms : int 2 1 1 1 1 3 2 1 1 1 ...
## $ bathrooms : num 1.5 1.5 1 1 1 1 1 1 1 1 ...
## $ number_of_reviews: int 29 0 73 2 0 14 248 44 0 0 ...
summary(df)
## price room_type city accommodates
## Min. : 1.0 Entire home/apt:28966 Boston : 2432 Min. : 1.00
## 1st Qu.: 75.0 Private room :21423 Chicago: 2573 1st Qu.: 2.00
## Median : 111.0 Shared room : 1490 DC : 3947 Median : 2.00
## Mean : 160.4 LA :15733 Mean : 3.15
## 3rd Qu.: 185.0 NYC :22655 3rd Qu.: 4.00
## Max. :1999.0 SF : 4539 Max. :16.00
## bedrooms bathrooms number_of_reviews
## Min. : 0.000 Min. :0.000 Min. : 0.00
## 1st Qu.: 1.000 1st Qu.:1.000 1st Qu.: 1.00
## Median : 1.000 Median :1.000 Median : 6.00
## Mean : 1.261 Mean :1.233 Mean : 20.82
## 3rd Qu.: 1.000 3rd Qu.:1.000 3rd Qu.: 23.00
## Max. :10.000 Max. :8.000 Max. :605.00
# PEMBAGIAN DATA TRAIN (80% training, 20% testing)
index <- sample(1:nrow(df), round(0.8 * nrow(df)))
train_data <- df[index, ]
test_data <- df[-index, ]
nrow(train_data)
## [1] 41503
nrow(test_data)
## [1] 10376
sd(test_data$price)
## [1] 168.2594
# MODEL GRADIENT BOOSTING
model_GBR <- gbm(
formula = price ~ .,
distribution = "gaussian", # loss kuadrat: 1/2 (y - F)^2
data = train_data,
n.trees = 1000, # batas atas jumlah pohon (M)
shrinkage = 0.05, # learning rate (v)
interaction.depth = 3, # kedalaman tiap pohon
n.minobsinnode = 10, # minimum observasi per daun
bag.fraction = 0.8, # tiap pohon memakai 80% data
cv.folds = 5, # 5-fold cross validation
verbose = FALSE
)
model_GBR
## gbm(formula = price ~ ., distribution = "gaussian", data = train_data,
## n.trees = 1000, interaction.depth = 3, n.minobsinnode = 10,
## shrinkage = 0.05, bag.fraction = 0.8, cv.folds = 5, verbose = FALSE)
## A gradient boosted model with gaussian loss function.
## 1000 iterations were performed.
## The best cross-validation iteration was 441.
## There were 6 predictors of which 6 had non-zero influence.
# Jumlah pohon optimal (early stopping via cross validation)
best <- gbm.perf(model_GBR, method = "cv")
best
## [1] 441
# PENTINGNYA VARIABEL (relative influence)
par(mar = c(5, 10, 2, 2))
imp <- summary(model_GBR, n.trees = best, plotit = TRUE, las = 1)
imp
## var rel.inf
## bedrooms bedrooms 28.882400
## bathrooms bathrooms 24.326745
## accommodates accommodates 17.382558
## room_type room_type 12.467774
## city city 8.563438
## number_of_reviews number_of_reviews 8.377085
# EFEK MARGINAL (partial dependence)
plot(model_GBR, i.var = "bedrooms", n.trees = best)
plot(model_GBR, i.var = "bathrooms", n.trees = best)
plot(model_GBR, i.var = "accommodates", n.trees = best)
plot(model_GBR, i.var = "room_type", n.trees = best)
plot(model_GBR, i.var = "city", n.trees = best)
plot(model_GBR, i.var = "number_of_reviews", n.trees = best)
# PREDIKSI DAN EVALUASI PADA DATA TESTING
pred_GBR <- predict(model_GBR, newdata = test_data, n.trees = best)
metrik <- function(aktual, prediksi) {
c(RMSE = sqrt(mean((aktual - prediksi)^2)),
MAE = mean(abs(aktual - prediksi)),
R2 = 1 - sum((aktual - prediksi)^2) / sum((aktual - mean(aktual))^2))
}
metrik(test_data$price, pred_GBR)
## RMSE MAE R2
## 118.4304426 62.0686399 0.5045382
# Lihat hasil prediksi
hasil <- test_data
hasil$prediksi <- pred_GBR
tail(hasil, 10)
## price room_type city accommodates bedrooms bathrooms
## 51824 100 Private room NYC 2 1 1
## 51826 119 Entire home/apt LA 1 0 1
## 51833 50 Private room NYC 2 1 1
## 51855 65 Private room LA 2 1 1
## 51865 75 Private room NYC 2 1 1
## 51866 150 Entire home/apt NYC 2 0 1
## 51870 95 Entire home/apt LA 4 0 1
## 51872 100 Private room NYC 2 1 1
## 51877 185 Entire home/apt NYC 5 2 1
## 51879 128 Entire home/apt LA 4 1 1
## number_of_reviews prediksi
## 51824 55 81.93235
## 51826 31 98.33694
## 51833 0 102.95077
## 51855 26 70.13682
## 51865 16 83.89093
## 51866 5 147.55183
## 51870 177 117.05415
## 51872 32 83.51347
## 51877 43 202.38698
## 51879 205 127.99602
# GRAFIK AKTUAL VS PREDIKSI (30 listing pertama data testing)
grafik <- data.frame(index = 1:30,
Aktual = hasil$price[1:30],
Prediksi = hasil$prediksi[1:30])
ggplot(grafik, aes(x = index)) +
geom_line(aes(y = Aktual, color = "Aktual")) +
geom_line(aes(y = Prediksi, color = "Prediksi")) +
geom_point(aes(y = Aktual)) +
geom_point(aes(y = Prediksi)) +
labs(x = "index", y = "price", color = "Keterangan:")
# TUNING HYPERPARAMETER DENGAN CARET (grid search, 5-fold CV)
ctrl <- trainControl(method = "cv", number = 5)
grid <- expand.grid(n.trees = c(50, 100, 200),
interaction.depth = c(2, 3, 4),
shrinkage = c(0.01, 0.05, 0.1),
n.minobsinnode = 10)
model_cv <- train(price ~ ., data = train_data,
method = "gbm",
trControl = ctrl,
tuneGrid = grid,
verbose = FALSE)
model_cv$bestTune
## n.trees interaction.depth shrinkage n.minobsinnode
## 27 200 4 0.1 10
head(model_cv$results[order(model_cv$results$RMSE), ])
## shrinkage interaction.depth n.minobsinnode n.trees RMSE Rsquared
## 27 0.10 4 10 200 119.4655 0.5009168
## 26 0.10 4 10 100 119.5967 0.5001003
## 24 0.10 3 10 200 119.6760 0.4993670
## 18 0.05 4 10 200 119.7780 0.4986695
## 23 0.10 3 10 100 120.2048 0.4953328
## 15 0.05 3 10 200 120.3682 0.4939011
## MAE RMSESD RsquaredSD MAESD
## 27 62.22150 1.494024 0.01789341 0.4392205
## 26 62.47220 1.477695 0.01857675 0.3548022
## 24 62.56417 1.380440 0.01699232 0.4580608
## 18 62.55562 1.310189 0.01752277 0.3868770
## 23 62.95249 1.216403 0.01672065 0.3893082
## 15 62.91205 1.599969 0.02003256 0.4112791
# MODEL DENGAN HYPERPARAMETER HASIL TUNING
terbaik <- model_cv$bestTune
model_tuning <- gbm(
formula = price ~ .,
distribution = "gaussian",
data = train_data,
n.trees = terbaik$n.trees,
shrinkage = terbaik$shrinkage,
interaction.depth = terbaik$interaction.depth,
n.minobsinnode = terbaik$n.minobsinnode,
bag.fraction = 0.8,
verbose = FALSE
)
model_tuning
## gbm(formula = price ~ ., distribution = "gaussian", data = train_data,
## n.trees = terbaik$n.trees, interaction.depth = terbaik$interaction.depth,
## n.minobsinnode = terbaik$n.minobsinnode, shrinkage = terbaik$shrinkage,
## bag.fraction = 0.8, verbose = FALSE)
## A gradient boosted model with gaussian loss function.
## 200 iterations were performed.
## There were 6 predictors of which 6 had non-zero influence.
pred_tuning <- predict(model_tuning, newdata = test_data,
n.trees = terbaik$n.trees)
# Perbandingan model utama dan model hasil tuning
rbind("Model Utama" = metrik(test_data$price, pred_GBR),
"Model Hasil Tuning" = metrik(test_data$price, pred_tuning))
## RMSE MAE R2
## Model Utama 118.4304 62.06864 0.5045382
## Model Hasil Tuning 118.2787 61.97061 0.5058067
# PREDIKSI HARGA UNTUK DATA BARU
data_test$prediksi_price <- predict(model_tuning, newdata = data_test,
n.trees = best)
## Warning in predict.gbm(model_tuning, newdata = data_test, n.trees = best):
## Number of trees not specified or exceeded number fit so far. Using 200.
head(data_test, 10)
## room_type city accommodates bedrooms bathrooms number_of_reviews
## 1 Entire home/apt LA 4 2 1.5 29
## 2 Private room Chicago 2 1 1.5 0
## 3 Entire home/apt LA 4 1 1.0 73
## 4 Private room LA 2 1 1.0 2
## 5 Entire home/apt DC 4 1 1.0 0
## 6 Entire home/apt NYC 6 3 1.0 14
## 7 Entire home/apt LA 4 2 1.0 248
## 8 Private room LA 8 1 1.0 44
## 9 Entire home/apt NYC 2 1 1.0 0
## 10 Private room NYC 1 1 1.0 0
## prediksi_price
## 1 191.48228
## 2 91.91072
## 3 129.83958
## 4 74.97097
## 5 396.67526
## 6 256.66769
## 7 150.61667
## 8 133.44691
## 9 185.19097
## 10 86.14146