Dear researchers,
I am wandering to know what are the disadvantages of CART-based prediction model in contrast to other models like nonparametric prediction method k-NN and the classic parametric method Kalman filter model.
I would be appreciative if you let me know your opinions.
Sincerely yours,
Mahdieh Zabihimayvan
Hi Mahdieh,
First of all, you will get more useful answers if you explain what you are using CARTs for!
At first sight I would say CARTs are very flexible and interpretable, however you must fully grasp how it works in order to set and cross-validate appropriate tuning parameters, such as tree depth / split number. Random forests (essentially collections of different CARTs) are much more user-friendly and robust compared to CARTs, why don't you give it a try? I personally think KNN is too simple and only advisable when no other method can handle bazillions of data. Cheers,
Francisco
- Badges
- Science topic