library(caret)

library(ISLR)

library(caTools)

train_data <- read.csv('C:\\Users\\Yevhen\\Documents\\Courses\\Data Mining\\HW6\\train.csv', sep=',', header=TRUE)
test_data <- read.csv('C:\\Users\\Yevhen\\Documents\\Courses\\Data Mining\\HW6\\test.csv', sep=',', header=TRUE)
set.seed(101)

split = sample.split(train_data$agegroup, SplitRatio = 0.90)
train = subset(train_data, split == TRUE)
test = subset(train_data, split == FALSE)
feats <- names(train[, 1:ncol(train)-1])
f <- paste(feats,collapse=' + ')
f <- paste('target ~',f)
f <- as.formula(f)

# prepare training scheme
control <- trainControl(method="repeatedcv", number=5, repeats=3)
# train the GBM model
set.seed(7)
modelGbm <- train(f, data=train, method="gbm", trControl=control, verbose=FALSE)
predvals <- predict(modelGbm, test[,1:ncol(test)-1])

n_test = nrow(test)
rmse_ = sqrt(sum((test$target-predvals)^2) / n_test)
rmse_


modelGbm <- train(f, data=train, method="gbm", trControl=control, verbose=FALSE)
test_pred_values <- as.data.frame(predict(modelGbm, newdata=test_data))
predict <- data.frame(test_data$ID, test_pred_values)
colnames(predict) <- c("ID", "target")
write.csv(predict, file = "C:\\Users\\Yevhen\\Documents\\Courses\\Data Mining\\prediction.csv", row.names=FALSE)