Global Induction of Oblique Model Trees: An Evolutionary Approach

Abstract

In this paper we propose a new evolutionary algorithm for global induction of oblique model trees that associates leaves with multiple linear regression models. In contrast to the typical top-down approaches it globally searches for the best tree structure, splitting hyper-planes in internal nodes and models in the leaves. The general structure of proposed solution follows a typical framework of evolutionary algorithms with an unstructured population and a generational selection. We propose specialized genetic operators to mutate and cross-over individuals (trees). The fitness function is based on the Bayesian Information Criterion. In preliminary experimental evaluation we show the impact of the tree representation on solving different prediction problems.