## Using GBM in R 
 
# install “gbm” & "readr"
library(readr)
library(gbm)

 
  train <- read_csv("../input/training.csv")
 
 str(train)

# The terms below are omitted to avoid failing the agreement test
# Remove some columns to pass the tests

train[,"id"] <- NULL
train[,"IPSig"] <- NULL
train[,"mass"] <- NULL
train[,"production"] <- NULL
train[,"min_ANNmuon"] <- NULL
train[,"SPDhits"] <- NULL

summary(train)

# Run GMB Generalized Boosted Regression Models

m <- gbm(signal ~ . , distribution="bernoulli", data=train, n.trees=500, interaction.depth=6, n.minobsinnode=10, shrinkage=0.1, bag.fraction=0.5, train.fraction=1, verbose=TRUE)

summary(m)

# load test file 
test <- read_csv("../input/test.csv")

# The terms below are omitted to avoid failing the agreement test
# Remove some columns to pass the tests

test[,"mass"] <- NULL
test[,"IPSig"] <- NULL
test[,"production"] <- NULL
test[,"min_ANNmuon"] <- NULL
test[,"SPDhits"] <- NULL

# make prediction and write submission

 yp <- predict(m, type="response", newdata=test, n.trees=300)
 
 sub <- data.frame(id=test$id, prediction=yp)
 summary(sub)

 write.csv(sub, "submissionGBM.csv", quote=F, row.names=F)