SVM Partitioned Decision Trees

This paper investigates the use of decision trees to improve the performance of a linear support vector machine. Specifically, it creates a decision tree and uses the results of an SVM prediction as a branching rule for the tree. When tested on a suit of 8 data sets, this hybrid methodology was seen to outperform the usual linear SVM on 3 of the data sets tested and in the other situations the performance of the two models were similar.