функция для кроссвалидации временных рядов, cv выборка для временных рядов с помощью TimeSeriesSplit

def cv_time_series_rmse(regressor, data, target_column, n_splits, max_train_size):
    tscv = TimeSeriesSplit(n_splits = n_splits, max_train_size = max_train_size)
    RMSE = 0.0
    for train_index, test_index in tscv.split(data):
        train = data.iloc[train_index]
        test = data.iloc[test_index]
        train_features = train.drop(target_column, axis=1)
        train_target = train[target_column]
        test_features = test.drop(target_column, axis=1)
        test_target = test[target_column]
        regressor.fit(train_features, train_target)
        RMSE += mean_squared_error(test_target, regressor.predict(test_features)) ** 0.5
    return RMSE / n_splits

cv_train_test_split = (train.shape[0] * 4 //5)
cv_splits = 3
cv_time_series_rmse(regressor=LinearRegression(), 
                    data=train, 
                    target_column='num_orders', 
                    n_splits=cv_splits, 
                    max_train_size=cv_train_test_split)