The Hybrid Dynamic Stock Forecasting Model Based on ANN and SVR

Lately, mass hybrid time series models have yet been applied to settle complexity nonlinear and financial issues. Nevertheless, some hybrid models have limitations, and the accuracy of separate forecasting model needs to be improved. Therefore, this paper demonstrates a novel hybrid dynamic stock time series forecasting model depended on ANN and SVR, which can be used to avoid that disadvantages mentioned hereinbefore. First, the appropriate data preprocessing is carried out to improve the quality of the feature. Second, it applies ANN and SVR to forecast the fluctuation of stock. Third, it improves the accuracy of prediction by dynamic adjustment bias. To evaluate the proposed hybrid and dynamic model, we used China stock exchange data as the datasets which from June 8, 2015 to May 26, 2016. The results clearly display that the offered model makes result reach a more precise prediction, and the accuracy can be achieved 79%.