data_manipulation.py 文件源码

def polynomial_features(X, degree):
    n_samples, n_features = np.shape(X)

    def index_combinations():
        combs = [combinations_with_replacement(range(n_features), i) for i in range(0, degree + 1)]
        flat_combs = [item for sublist in combs for item in sublist]
        return flat_combs

    combinations = index_combinations()
    n_output_features = len(combinations)
    X_new = np.empty((n_samples, n_output_features))

    for i, index_combs in enumerate(combinations):  
        X_new[:, i] = np.prod(X[:, index_combs], axis=1)

    return X_new