xgboost_fastsetup.py

xgb_params = {'objective': 'multi:softmax',
              'num_class': 19,
              'tree_method': 'exact',
#                      'max_bin': 64,
              'colsample_bytree': 0.75,
              'subsample': 0.75,
#               'lambda': 2,
#               'alpha': 2,
#               'min_child_weight': 10,
#               'max_delta_step': 10,
              'max_depth': 10,
              'eval_metric': ['mlogloss'],
              'learning_rate': 0.03}


cv_hist = xgb.cv(params=xgb_params,
                 dtrain=dm_train,
                 verbose_eval=25,
                 folds=email_folds,
                 num_boost_round=10000,
                 early_stopping_rounds=100)


best_iter = np.argmin(cv_hist['test-mlogloss-mean'])
gbm_model = xgb.train(xgb_params,
                      dm_train,
                      verbose_eval=100,
                      num_boost_round=best_iter)