Research On The Prediction Of The Price In SSE 50 Stock Based On Support Vector Regression

With the rapid development of Chinese stock market,a large amount of stock data will be generated every day.Effective prediction of these data is a very important and valuable issue for stock investors.In this paper,the SSE 50 index is the research object.Based on the extensive review of the literature related to the stock price forecasting method,a support vector regression model is used to forecast the stock price.First,through the correlation test,the six indicators commonly used in the SSE 50 Index are selected as independent variables,and the opening price of the second day is used as the dependent variable.Five models were established from historical data from the previous day to the first five days.The model uses support vector regression to train and predict,and selects the best prediction model for the opening price of the second day based on model error predictions.However,due to the influence of multiple factors,it is often impossible to accurately predict the second day opening price Therefore,we propose to use the time window as an independent variable to blur-granulate the opening price,predict the opening price of the opening price and the range of variation through fuzzy particles,and use support vector regression to train and forecast.The results show that the prediction error of the model is larger in the stock fluctuation range,so we propose an improvement to the model,select the fuzzy grained indicator as an independent variable,and eliminate the influence of time on model prediction.The results of the study found that the six indicators of the previous day had the best effect in predicting the opening price of the second day,which is different from the input variables we commonly think of using as much as possible;in the forecast of the change trend and scope of the opening price,through the improvement of the support vector regression model,we basically eliminate the impact of stock price fluctuation on the forecast error and improve the prediction accuracy of the model.