epl_engine.py

import os.path
import pickle
import time
from typing import Any
import numpy as np


SAVE_DIR = 'models/trained'
os.makedirs(SAVE_DIR, exist_ok=True)


def train(clf: Any, X_train: np.ndarray, y_train: np.ndarray) -> Any:
    start = time.time()
    clf.fit(X_train, y_train)
    print('Training took {:.04f}secs'.format(time.time() - start))
    return clf


def test(clf: Any, X_test: np.ndarray, y_test: np.ndarray) -> float:
    accuracy = clf.score(X_test, y_test)
    return accuracy


def predict(clf, X: tuple[np.ndarray, np.ndarray]) -> np.ndarray:
    return clf.predict(X)


def train_predict(
    clf: Any,
    X_train: np.ndarray,
    y_train: np.ndarray,
    X: tuple[np.ndarray, np.ndarray],
):
    train(clf, X_train, y_train)
    return predict(clf, X)


def save_classifier(
    clf: Any,
    filename: str | None = None,
    force: bool = False
) -> None:
    if not filename:
        model_name = str(clf)
        filename = model_name[:model_name.index('(')]
    filename = os.path.join(SAVE_DIR, '{}.pkl'.format(filename))
    if force or not os.path.exists(filename):
        f = open(filename, 'wb')
        pickle.dump(clf, f)
        f.close()
        print(f'Successfully saved to {filename}!')
    else:
        raise (Exception('File already exist!'))


def main() -> None:
    from sklearn.ensemble import AdaBoostClassifier
    from .preprocess import process, process_to_features

    # Load all the datasets
    home_team = 'arsenal'
    away_team = 'stoke'

    X_train, X_test, y_train, y_test = process(
        filename=None, test_size=0.1, save_csv=True
    )
    pred_features = process_to_features(home=home_team, away=away_team)

    print('Training: ', X_train.shape, y_train.shape)
    print('Testing:  ', X_test.shape, y_test.shape)

    try:
        # !- TODO: Try a bunch of classifiers to get better accuracy
        clf = AdaBoostClassifier(n_estimators=500, learning_rate=1e-2)
        train(clf, X_train, y_train)
        accuracy = test(clf, X_test, y_test)
        pred = predict(clf, pred_features)

        print('Soccer prediction accuracy = {:.02%}\n'.format(accuracy))
        print('Prediction: ', pred)

        save_classifier(clf, force=True)
    except Exception as e:
        import sys

        sys.stderr.write(str(e))
        sys.stderr.flush()


if __name__ == '__main__':
    main()