gistfile1.txt

param <- list(  objective           = "binary:logistic", 
                booster             = "gbtree",
                eval_metric         = "auc",  # maximizing for auc
                eta                 = 0.02,   # learning rate - Number of Trees
                max_depth           = 5,      # maximum depth of a tree
                subsample           = .9,     # subsample ratio of the training instance
                colsample_bytree    = .87,    # subsample ratio of columns 
                min_child_weight    = 1,      # minimum sum of instance weight (defualt)
                scale_pos_weight    = 1       # helps convergance bc dataset is unbalanced
                ) 
                
xgb <- xgb.train(   params              = param, 
                    data                = dtrain, 
                    nrounds             = 750, 
                    verbose             = 1,
                    watchlist           = watchlist,
                    maximize            = FALSE
                    )