Induction of Oblique Decision Trees

This paper introduces a randomized technique for partitioning examples using oblique hyperplanes. Standard decision tree techniques , such as ID3 and its descendants, partition a set of points with axis-parallel hyper-planes. Our method, by contrast, attempts to nd hyperplanes at any orientation. The purpose of this more general technique is to nd smaller but equally accurate decision trees than those created by other methods. We have tested our algorithmon both real and simulated data, and found that in some cases it produces surprisingly small trees without losing predic-tive accuracy. Small trees allow us, in turn, to obtain simple qualitative descriptions of each problem domain.