clf_for_eval.set_params(**clf_gs.best_params_)
metrics, std = training_utils.evaluate_on_cv(training_data,
train_X_all,
clf_for_eval,
fuzzy_option,
fuzzy_dist_column,
fuzzy_err_column,
xgb_flag=True)
pr_curve = training_utils.predict_and_pr_curve_on_cv(training_data,
train_X_all,
clf_for_eval,
fuzzy_option,
fuzzy_dist_column,
fuzzy_err_column,
xgb_flag=True)
# best model from grid search:
clf_best = clf_gs.best_estimator_
# generalization:
general_data["y_pred"] = clf_best.predict(general_X)
general_data["y_prob_positive_class"] = clf_best.predict_proba(
general_X)[:, 1]
training_utils.save_results(output_path, experiment_name, fuzzy_option,
training_data, general_data, metrics, std,
pr_curve, best_param_df, gs_results_df,
info_columns, features)
print("done.")
fuzzy_dist_column,
fuzzy_err_column)
pr_curve = training_utils.predict_and_pr_curve_on_cv(
training_data, train_X, clf, fuzzy_option, fuzzy_dist_column,
fuzzy_err_column)
# fit to the data:
if fuzzy_option == "normal":
clf.fit(X=train_X, y=training_data["Y"])
elif fuzzy_option == "fuzzy_dist":
clf.fit(X=train_X,
y=training_data["Y"],
sample_weight=training_data[fuzzy_dist_column].values.T[0])
elif fuzzy_option == "fuzzy_err":
clf.fit(X=train_X,
y=training_data["Y"],
sample_weight=training_data[fuzzy_err_column].values.T[0])
else:
print("wrong fuzzy option")
# generalization:
general_data["y_pred"] = clf.predict(general_X)
general_data["y_prob_positive_class"] = clf.predict_proba(general_X)[:, 1]
training_utils.save_results(output_path, experiment_name, fuzzy_option,
training_data, general_data,
metrics, std, pr_curve, pd.DataFrame(),
pd.DataFrame(), info_columns, features)
print("done.")
