Geometric Approach for Induction of Oblique Decision Tree

Abstract-In this paper we present new algorithm for oblique deision tree induction. We propose new classifier that performs better than the other decision tree approaches in terms of accuracy, size, time. Proposed algorithm uses geometric structure in the data for assessing the hyper planes. At each node of the decision tree, we suggest the clustering hyper planes for both the classes and using this representation their angle bisectors is selected as split rule at that node. The algorithm we present here is applicable for2-class and multiclass problems. Through empirical investigation we demonstrate that this idea leads to small decision trees and better performance. We also present some analysis to show that the angle bisectors of clustering hyperplanes that we use as the split rules at each node are solutions of an interesting optimization problem and hence argue that classifier obtained with new approach is as good and novel classification method. Keywords- oblique decision tree,CART,OC1,SVM,GDT I