8.Performance comparison of Regression Model, Regression Tree, Bagging Tree, Random Forest and Gradient Boosted Model.Rmd

---
output:
  word_document: default
  html_document: default
---
```{r}
data <- palmerpenguins::penguins
anyNA(data)
dim(data)
data <- na.omit(data)
anyNA(data)
dim(data)
```



DATA STRUCTURE
```{r}
head(data)
str(data)
summary(data)
```



SPLITTING DATA

```{r}
#Splitting Data
library(caret)
set.seed(2380)
index <- createDataPartition(data$body_mass_g, p = 0.75,
list = FALSE, times = 1)
train <- data[index,]
test <- data[-index,]
nrow(train)
nrow(test)
```



LINEAR REGRESSİON MODEL

```{r}
#Traing the Model
model_LM <- lm(body_mass_g~. ,data=train)
#Performance for Train
predicted_train_LM <- predict(model_LM, train)
rmse_train_LM <-sqrt(mean((predicted_train_LM - train$body_mass_g) ^ 2))

#Performance for Test
predicted_test_LM <- predict(model_LM, test)
rmse_test_LM <- sqrt(mean((predicted_test_LM - test$body_mass_g) ^ 2))

cat("test_rmse:", rmse_test_LM,"\n")
cat("train_rmse:", rmse_train_LM)
```


REGRESSION TREE

```{r}
#Training the Model
library(rpart)
library(rpart.plot)
model_DT <- rpart(body_mass_g ~. , method = "anova", data = train)

#Performance for test
predicted_test_DT <- predict(model_DT, test)
rmse_test_DT <- sqrt(mean((predicted_test_DT - test$body_mass_g) ^ 2))

#Performance for train
predicted_train_DT <- predict(model_DT, train)
rmse_train_DT <- sqrt(mean((predicted_train_DT - train$body_mass_g) ^ 2))

cat("test_rmse:", rmse_test_DT,"\n")
cat("train_rmse:", rmse_train_DT)
```



BEGGING TREE
```{r}
#Training the Model
set.seed(2380)
library(randomForest)
model_BT <- randomForest(body_mass_g~. , data=train, mtry= 7)
plot(model_BT,col="red")
which.min(model_BT$mse)
#Performance for test
predicted_test_BT <- predict(model_BT, test)
rmse_test_BT <- sqrt(mean((predicted_test_BT - test$body_mass_g) ^ 2))

#Performance for train
predicted_train_BT <- predict(model_BT, train)
rmse_train_BT <- sqrt(mean((predicted_train_BT - train$body_mass_g) ^ 2))

cat("test_rmse:", rmse_test_BT,"\n")
cat("train_rmse:", rmse_train_BT)

```


RANDOM FOREST

```{r}
#Training Model
set.seed(2380)
model_RF <- randomForest(body_mass_g ~ ., data = train)
plot(model_RF,col  = "blue")
which.min(model_RF$mse)
#Performance for test
predicted_test_RF <- predict(model_RF, test)
rmse_test_RF <- sqrt(mean((predicted_test_RF - test$body_mass_g) ^ 2))

#Performance for train
predicted_train_RF <- predict(model_RF, train)
rmse_train_RF <- sqrt(mean((predicted_train_RF - train$body_mass_g) ^ 2))

cat("test_rmse:", rmse_test_RF,"\n")
cat("train_rmse:", rmse_train_RF)
```




GRADIENT BOOSTED MODEL

```{r}
library(gbm)
#Training Model
set.seed(2380)
model_GBM <- gbm(body_mass_g ~ ., data = train, distribution = "gaussian", n.trees = 100, cv.folds = 5)
gbm.perf(model_GBM)
#Performance for test
predicted_test_GBM <- predict(model_GBM, test)
rmse_test_GBM <- sqrt(mean((predicted_test_GBM - test$body_mass_g) ^ 2))

#Performance for train
predicted_train_GBM <- predict(model_GBM, train)
rmse_train_GBM <- sqrt(mean((predicted_train_GBM - train$body_mass_g) ^ 2))

cat("test_rmse:", rmse_test_GBM,"\n")
cat("train_rmse:", rmse_train_GBM)
```







```{r}
#Comparing models performances
data.frame(
  "TEST" = c(rmse_test_BT,rmse_test_DT,rmse_test_LM,rmse_test_RF,rmse_test_GBM),
  "TRAIN" = c(rmse_train_BT,rmse_train_DT,rmse_train_LM,rmse_train_RF,rmse_train_GBM),
  row.names = c("BEGGING TREE", "DECISION TREE","LINEAR MODEL","RANDOM FOREST","GRADIENT BOOSTED")
)
```