A COMPARISON OF DECISION TREE CLASSIFIERS WITH BACKPROPAGATION NEURALNETWORKS FOR MULTIMODAL CLASSIFICATION PROBLEMS

Multi-modal classification problems involve the recognition of pattern s where the patterns associated with each class can come from disjoint regions in feature space. Traditional linear discriminant methods can not cope with these problems. While a number of approaches exist for c lassifying patterns with multiple modes, decision trees and backpropag ation neural networks represent leading algorithms with special capabi lities for dealing with this problem class. This paper provides a comp arison of decision trees with backpropagation neural networks for thre e distinct multi-modal problems: two from emitter classification and o ne from digit recognition. These real-world problems provide an intere sting range of problem characteristics for our comparison: one emitter classification problem has few features and a large data set; and the other has many features and a small data set. Additionally, both emit ter classification problems have real-valued features, while the digit recognition problem has binary-valued features. The results show that both methods produce comparable error rates but that direct applicati on of either method will not necessarily produce the lowest error rate . In particular, we improve decision tree results with multi-variable splits and we improve backpropagation neural networks with feature sel ection and mode identification.