utils.py 文件源码

def compute_loss(input_compute_loss):

    Model = input_compute_loss["Model"]
    config = input_compute_loss["config"]
    X_train = input_compute_loss["X_train"]
    y_train = input_compute_loss["y_train"]
    dates_train = input_compute_loss["dates_train"]
    X_test = input_compute_loss["X_test"]
    y_test = input_compute_loss["y_test"]
    is_y_log = input_compute_loss["is_y_log"]
    is_boxcox = input_compute_loss["is_boxcox"]
    loss_func = input_compute_loss["loss_func"]

    model = Model(**config)
    if hasattr(model ,"dates_train"):
        model.dates_train = dates_train
    if is_y_log:
        model.fit(X_train, np.log(y_train))
        predict_y_test = np.exp(model.predict(X_test))
    elif is_boxcox:
        model.fit(X_train, boxcox(y_train, boxcox_lambda))
        predict_y_test = invboxcox(model.predict(X_test), boxcox_lambda)
    else:
        model.fit(X_train, y_train)
        predict_y_test = model.predict(X_test)
    if loss_func is None:
        loss = mape_loss(y_test, predict_y_test)
    else:
        loss = loss_func(y_test, predict_y_test)
    return (repr(config), config, loss)