Compare mean absolute error (MAE) values according to internal and external validations of various machine learning models.