Abstract
Features from the Saudi Stock Market (SSM) have been examined to attempt to predict the direction of daily price changes. Backpropagation neural network has been applied to predict the direction of price changes for the listed stocks in SSM. The price change in SSM ranges between -10% and 10%. The target has a representation of three classes 1, -1 and 0 that respectively represent the increase, decrease or insignificant change in the stock prices. The dynamic target is a novel enhancement to the traditional objective function mean-squared-error (MSE) for better classification. Our preliminary results show that the classifier's performance improved using dynamic targets in terms of quantitative performance and qualitative performance. In addition, experiments were conducted to determine the best hardening function for objective targets.