The Efficacy of Neural Networks and Simple Technical Indicators in Predicting Stock Markets

This paper investigates the efficacy of neural networks and simple technical indicators in predicting stock market movement. The prediction system uses a back-propagation neural network and the KD and %R indicators. Our results show that monthly indicators respond too slowly to effectively capture the market trends. The %R indicator is a better market predictor than the KD indicator when they are used alone. The daily %R, weekly %R, weekly KD indicators, and their combinations can provide reasonable predictions with a percentage of correct predictions of around 60%. If the predictions of sideway-movements are excluded, the prediction accuracy can increase to about 80% Our neural network prediction system works equally well on both the TSE market and the Nasdaq market. Though specialized for the KD and %R indicators, many aspects of this methodology can be generalized to check the validity of other technical indicators.