Forecasting Stock Market Indices with Recurrent Neural Networks

Abstract

A recurrent neural network is used to forecast the out-of-sample return of a stock market index. The use of an extensive information set and a stochastic minimization algorithm distinguishes this study from prior work. The data set encompasses daily observations from 1970 through 1993, with the following forecast exercise undertaken. For a variety of model sizes, the network task is to approximate the weekly, monthly or quarterly conditional mean return. These forecasts are conditioned on a daily information set containing a number of index-specific and market-wide variables, term structure and corporate bond yields, and calendar variables. Network performance is evaluated by out-of-sample normalized mean-squared error, sample statistics describing the joint distribution of forecasted and actual returns, and a test for market-timing ability. A further performance evaluation concerns the construction of trading portfolios with transaction costs. Finally, bootstrapping techniques are applied to construct surrogate distributions of the out-of-sample statistics. Neural network models are found to perform more than adequately when compared with a benchmark linear model, and are able to generate large risk-adjusted returns over simple buy-and-hold strategies.

This chapter is based on work contained in my doctoral dissertation. A version of this paper was presented at the Second International Workshop on Neural Networks in the Capital Markets, Pasadena, CA, November 1994. I wish to thank Masanao Aoki for introducing me to this topic and for his guidance and criticism.