max_f1 = 0. # initializing with 0
best_params = None
for max_depth, min_child_weight in gridsearch_params:
print("CV with max_depth={}, min_child_weight={}".format(
max_depth,
min_child_weight))
# Update our parameters
params['max_depth'] = max_depth
params['min_child_weight'] = min_child_weight
# Cross-validation
cv_results = xgb.cv(params, dtrain, feval= custom_eval, num_boost_round=200,
maximize=True,
seed=16,
nfold=5,
early_stopping_rounds=10
)
# Finding best F1 Score
mean_f1 = cv_results['test-f1_score-mean'].max()
boost_rounds = cv_results['test-f1_score-mean'].argmax()
print("\tF1 Score {} for {} rounds".format(mean_f1, boost_rounds))
if mean_f1 > max_f1:
max_f1 = mean_f1
best_params = (max_depth,min_child_weight)
print("Best params: {}, {}, F1 Score: {}".format(best_params[0], best_params[1], max_f1))
d_train = xgb.DMatrix(Xtrain_hashed, ytrain)
d_valid = xgb.DMatrix(Xtest_hashed, ytest)
# XGBoost parameters 1
params = {
'n_estimators': 1024,
'n_jobs': -1,
'learning_rate': 1,
'eval_metric': 'auc'
}
# Without early stopping, overfit model
bst_cv = xgb.cv(params,
d_train,
num_boost_round=10000,
nfold=10,
verbose_eval=100,
early_stopping_rounds=None,
as_pandas=True)
bst_cv.tail()
bst_cv[['train-auc-mean', 'test-auc-mean']].plot()
print(bst.eval(d_valid))
# With early stopping. Use CV to find the best number of trees
bst_cv = xgb.cv(
params,
d_train,