Trading Equity Index Futures With a Neural Network

in La Jolla (CA 92037). eural network technology, which has been receiving increasing attention in the investN ment community, represents a radically different form of computation from the conventional algorithmic model. Neural networks consist of multiple simple processors arranged in a communicative network, each programmed to perform one identical, elementary processing task. This technology is especially suited for simulating intelligence in pattern detection, association, and classification activities. Such problems arise frequently in areas such as credit assessment, security investment, and financial forecasting, so it is not surprising that after the Department of Defense, which in 1989 embarked on a five-year, multi-million dollar program for neural network research, financial services organizations have been the principal sponsors of research in neural network applications. We describe here a specific neural networkbased day trading system for Standard & Poor’s (S&P) 500 index futures contracts that has, in ex ante evaluation, outperformed passive investment in the index. This system, which is fairly representative of a neural network-based trading strategy implementation, also demonstrates how performance can be enhanced by integrating neural networks with conventional rulebased expert system techniques (e.g., see Lee, Trippi, Chu, and Kim [1990] and Trippi [1990]).