FINALMENTEEEEEEEEEEEEEEEEEE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

gistfile1.txt

features = np.array([X**i for i in range(1, 11)]).T

best_rss_set = []
best_features_idx = []
best_cp = np.inf
best_model_features = []
# rever
full_model = LinearRegression().fit(features, Y)
sigma_hat_squared = mean_squared_error(Y, full_model.predict(features))

features_idx = [i for i in range(features.shape[1])]

while features_idx:
    best_rss = None
    best_idx = None
    for feature_idx in features_idx:
        features_set = features[:, best_features_idx + [feature_idx]]
        model = LinearRegression().fit(features_set,Y)
        y_pred = model.predict(features_set)
        Rss = np.sum((Y - y_pred)**2)

        if best_rss is None or Rss < best_rss:
            best_rss = Rss
            best_idx = feature_idx

    best_rss_set.append(best_rss)
    best_features_idx.append(best_idx)
    features_idx.remove(best_idx)

    p = len(best_features_idx) + 1

    # rever como é o cálculo do c_p statistics que o livro pede
    c_p = (best_rss / sigma_hat_squared) - (n - 2 * p)

    if c_p < best_cp:
        best_cp = c_p
        best_model_features = list(best_features_idx)
        final_model = LinearRegression().fit(features[:, best_features_idx], Y)
        coefficients = final_model.coef_

    print(f'O c_p statistics de {best_features_idx} é {c_p}')

print(f'\nO modelo selecionado de acordo com o c_p statistics é:{best_model_features}, e os coeficientes são: {coefficients}')