computationally inexpensive, capable in dealing with noisy data, its prediction model is
explainable to the user, it has automatic interaction detection to find the significant high-order interactions quickly, and it can produce more informative outputs [18, 20].