CORE - a system that predicts continuous variables

We describe CORE, an inductive system for learning regression trees. CORE incorporates some novel techniques for attribute estimation and constructive induction and can form various types of models in the leaves of the tree. Experimental results and its use in practice show that it is powerful tool for the analysis of the regression problems. It provides the user not only with the better predictive accuracy but also with the models that better capture the human understanding of the underlying phenomena.