Two-Stage Constructing Hyper-Plane for Each Test Node of Decision Tree
How to construct the “appropriate” split hyper-plane in test nodes is the key of building decision trees. Unlike a univariate decision tree, a multivariate (oblique) decision tree could find the hyper-plane that is not orthogonal to the features’ axes. In this paper, we re-explain the process of building test nodes in terms of geometry. Based on this, we propose a method of learning the hyper-plane with two stages. The tree (TSDT) induced in this way keeps the interpretability of univariate decision trees and the trait of multivariate decision trees which could find oblique hyper-plane. The tests of the impact of Combination methods tell us that TSDT based combination algorithm is much better than other tree based combination methods in accuracy.
Zhenyu Du and Bin Liu
W. She et al., "Two-Stage Constructing Hyper-Plane for Each Test Node of Decision Tree", Applied Mechanics and Materials, Vols. 26-28, pp. 776-779, 2010