# Otherwise, use default number try: train_test_split_folds = int(options["-t"]) except: train_test_split_folds = default_train_test_split_folds # We are using Stratified Shuffle Split, the same cross-validation method as # used in the tester shuffler = StratifiedShuffleSplit(labels, train_test_split_folds, random_state = 42) # Run through all the train-test splits, train, test and measure key metrics if gridsearchcv == "out": if len(search_grid)>0: # Specify CV option wrapped_algorithm.cv = shuffler # Fit classifier clf.fit(features, labels) # Get the best parameters of cross validation params = wrapped_algorithm.best_params_ # Convert them to a JSON object with every parameter in a # one-element array grid_json = json.dumps(params, ensure_ascii=True) # Write them to the csv file with open(gridsearchcv_file_name, 'ab') as csvfile: writer = csv.writer(csvfile, delimiter=",") writer.writerow([feature_scaling, feature_selection, clf_id, grid_json]) # Print clf pipeline on the screen so we see what the script is using. pprint.pprint(clf) # Print parameters we found