Stock Index Prediction Based On A Combined Model

As a representative statistical indicator of financial market, the fluctuation of stock price index not only reflects the market conditions and affects the interests of all investors, but also reflects the degree of market prosperity. Therefore, the forecasting analysis on time series of stock’s closing price attracts the attention of many research scholars. Besides, it has important theoretical and practical significance.On the basis of the current research including domestic and foreign, we find the shortcomings of single models and think the combined model is the main direction of future research. This paper selects monthly and semi-annual closing price data of Shanghai Composite Index, using the optimal matrix method to combine the most commonly used exponential smoothing model in economic forecasting, autoregressive conditional heteroscedasticity model and BP neural network model in order to achieve a more stable and higher prediction accuracy. The research focuses on the building of the combined model and comparing the empirical results of different models. And the final result shows that the combined model based on matrix method handles the information more efficiently and behaves better than other models.