boostedtreemodel

target = ifelse(train==0,"No","Yes")

#Create a small testing set of 1000 randomly selected training set observations
#against which to test the model being fitted

idx = sample(nrow(train),1000,replace = FALSE)

eval = train[idx,]
target.e = eval$target
master = train[-idx,]
target.m = master$target

#Necessary procedural steps to convert eval and target dataframes
#to a special XGBoost matrix for use in the train. train.xgb method

eval.f = sparse.model.matrix(target ~ ., data = eval)
train.f = sparse.model.matrix(target ~ ., data = master)
dval <-  xgb.DMatrix(data=eval.f, label=target.e)
dtrain <- xgb.DMatrix(data=train.f, label=target.m)
watchlist <- list(val=dval, train=dtrain)

#Model with Parameters

param <- list(  objective           = "binary:logistic", 
                booster             = "gbtree",
                eval_metric         = "logloss",
                eta                 = 0.01,
                max_depth           = 08,
                subsample           = .5,
                colsample_bytree    = 1,
                min_child_weight    = 1,
                num_parallel_tree   = 1
)

clf <- xgb.train(   params              = param, 
                    data                = dtrain, 
                    nrounds             = 2000, 
                    verbose             = 1,
                    watchlist           = watchlist,
                    early.stop.round    = 300,
                    maximize            = FALSE
)