Implementation or regression tree using roulette to choose the split

Authors

• Marcin Dadura
• Maciej Włodarczyk

Configuration

• tree.configure(roulette_option=True) - enable roulette, if false in function best_split() picked will be value with highest probability
• tree.configure(rounding_amount=4) - amount of rounding of values when printing tree
• tree.configure(width_print=9) - spaces padding when printing tree

Help functions

• dataFrame = prepare_data_frame(file_path=file_path, columns_name=np.concatenate((to_estimate_column_name, features_columns_name), axis=None), size=quantity_from_csv) - return datafram from file and parse this data
• X, Y = prepare_data(df=dataFrame, to_estimate=to_estimate_column_name, features=features_columns_name) - return X and Y vectors needed to generate regression tree
• error = compute_error(df=dataFrame, true_value=to_estimate_column_name, predicted=predicted_values_column_name - compute error

Usage

• tree = Node(X=X, Y=Y, max_depth=max_depth, min_elements=min_elements) - initialize root of tree
• tree.grow() - generate tree
• tree.print_tree() - print tree
• tree.predict(df=dataFrame, new_column_name=predicted_values_column_name) - predict estimated value, add new column with predicted value