Research On Prediction Of Stock Price Based On Combination Model

With our country's economic development and the continuous improvement of people's living standards,ways of investment and financial management are also constantly changing.In order to obtain greater returns,more and more people are beginning to pay attention to the stock investment market and participate in it.While investing in the stock market brings benefits to investors,there are also higher risks.Therefore,prejudgment of stock is particularly important.At present,the methods of stock price prediction generally use a single model and a combined model to predict.Many scholars usually use a single traditional time series model or neural network model and combine the models to make predictions.Empirical analysis shows that simply using time series models or neural network models to predict stock prices will bring greater errors.The appropriate combination model can improve the accuracy of prediction.Based on this,this paper selects the closing prices of the CSI 300 Index and the Shanghai Composite Index as the forecast objects,and establishes a suitable combination model to predict stock prices.First,use the ARIMA model and the BP neural network model as well as the principal component generalized regression neural network model PCA?GRNN to predict the closing prices of the two stock indexes,and compare the prediction accuracy of the three single models.It is found that the prediction effect of the PCA?GRNN model has the best prediction effect;Using three linear weighting methods,that is weight method,reciprocal variance method and dominance matrix method,three single models are combined in pairs to predict the closing price of two stock indexes.The results show that the PCA?GRNN model is the same as the ARIMA model and the ARIMA model.The combination of BP neural network model is better than ARIMA,BP neural network model and ARIMA?BP combined model,but its prediction accuracy is worse than single model PCA?GRNN;finally,based on the PCA?GRNN model,two residual optimization combined models are established,namely PCA?GRNN?ARIMA model and PCA?GRNN?GM(1,1)model.The results show that the prediction accuracy of the two residual optimization combination models are better than the PCA?GRNN model,especially the PCA?GRNN?ARIMA model which has the best prediction effect.