2015 Summer Symposium of Social Sciences and Society (Econf-SSS)

12 / 2015-06-18 17:11:42

A novel hybrid model based on EMD-BPNN for forecasting two stock markets

Stock price prediction is regarded as a challenging task of the financial time series prediction process. Recently, applying the novel data mining techniques for financial time series forecasting has received much research attention. Following empirical mode decomposition (EMD) and neural network theory, a method is presented to model and forecast stock market. First, using EMD theory, the stock market time series is decomposed into many intrinsic mode functions (IMF) which can significantly represent potential information of original time series, and the further analysis of IMF indicates that stock markets exist a chaos feature. Then, by using back propagation neural network (BPNN), the forecasting models are established to forecast the IMF respectively. By these means, the model can be improved to learn various objective functions and more precious prediction can be obtained. The novel hybrid model is based on the idea of decomposition-reconstruction-integration. In order to evaluate the forecasting performances, the proposed model is compared with ARIMA, GARCH and BPNN model. The experimental results show that the proposed model is superior to other models in terms of directional symmetry (DS).