Performance Comparisons Between Backpropagation Networks and Classification Trees on Three Real-World Applications

Part of Advances in Neural Information Processing Systems 2 (NIPS 1989)

Authors

Les Atlas, Ronald Cole, Jerome Connor, Mohamed El-Sharkawi, Robert Marks, Yeshwant Muthusamy, Etienne Barnard

Abstract

Multi-layer perceptrons and trained classification trees are two very different techniques which have recently become popular. Given enough data and time, both methods are capable of performing arbi(cid:173) trary non-linear classification. We first consider the important differences between multi-layer perceptrons and classification trees and conclude that there is not enough theoretical basis for the clear(cid:173) cut superiority of one technique over the other. For this reason, we performed a number of empirical tests on three real-world problems in power system load forecasting, power system security prediction, and speaker-independent vowel identification. In all cases, even for piecewise-linear trees, the multi-layer perceptron performed as well as or better than the trained classification trees.

Performance Comparisons

623