Hybrid learning fuzzy neural models in stock price forecasting

Abstract

Prediction of stock markets is one of the most important topics in business. However, the stock markets are very complicated and thus difficult to predict. The ellipsoidal fuzzy system (EFS) learning with and without supervision has been successfully applied in solving control systems and pattern recognition problems. This study modified the ellipsoidal fuzzy system to accept time series data in forecasting stock prices. Furthermore, a scaled conjugate gradient learning method was applied to accelerate supervised learning process. Finally, back-propagation neural networks (BPNN) are used to compare the forecasting accuracy. The empirical results including 10 companies show that the presented mode has higher prediction accuracy than the BPNN model.