Searching for Splitting Criteria in Multivariate Decision Tree Using Adapted JADE Optimization Algorithm
Chattriya Jariyavajee, Jumpol Polvichai, Booncharoen Sirinaovakul · 2019
Splitting criteria in the decision tree are normally univariate and orthogonal to the parameter axis. This limitation suppresses the decision power of DT and can be overcome by using multivariate splitting criteria. This paper presents an improvement of splitting criteria searching in a decision tree based on swarm intelligence, the global optimization technique. JADE, one of swarm intelligent algorithms, is adapted in this work to search for optimal parameters for the multivariate linear splitting criteria in the multivariate decision tree building. This paper illustrates the splitting criteria in 2D data space and tests the efficiency of the created decision tree. The experimental results show that AJADE-MDT, the proposed algorithm could be used to create the decision tree with higher efficiency compared to the existing algorithms.