Trained hyper-parameters

params = {
   'base_score': 0.5,
   'booster': 'gbtree',
   'colsample_bytree': 0.5,
   'eta': 0.5,
   'eval_metric':  ['error', 'mae', 'rmse'],
   'max_depth': 20,
   'nthread': 4,
   'objective': 'binary:logistic',
   'reg_alpha': 0.01,
   'reg_lambda': 0.1,
   'scale_pos_weight':9.85,
   'max_delta_step':1,
   'seed': 12345,
   'silent': 0,
   'subsample': 0.5
}


Number of trees: 110

Results evaluation:

best: [107] tr-error:0.021897	tr-mae:0.041504	tr-rmse:0.134437 val-error:0.061934	val-mae:0.100498	val-rmse:0.228714

trained ../mat_training.train mae: 0.0411 rmse: 0.1339
avg 0.092114 mae: 0.1673 rmse: 0.2892
Predicted correctly: 2033259/2078354


Stats data set:

Training:

nb_negatives: 1886908 is:90.78857595962959149400 nb_positives: 191446 is: 9.21142404037040850500 nb_examples 2078354


Test:

nb_negatives: 966977 is:94.29855613481796142500 nb_positives: 58465 is: 5.70144386518203857400 nb_examples 1025442