Research On Quantitative Investment Based On Data Analysis

In view of the complexity and uncertainty of stock price point prediction,the limitations of traditional statistical analysis methods,and the over-fitting of machine learning algorithms,this paper proposes the stock price range prediction model based on information granulation and BP-Bagging,ARIMA-IGS-SVM and verifies the validity of the model through quantitative analysis of CITIC Securities stocks.Information granulation technology can collect effective information according to the needs of research problems,so that it can represent the fuzzy features of the original data and is suitable for the analysis and research of stock price fluctuation range.Firstly,fuzzy information granulation is conducted on the original time series data to obtain the Low,R and Up values of the respective window.Secondly,considering the over-fitting and instability of BP neural network,the BP-Bagging prediction model is established.Experiments with R values show that Bagging algorithm can enhance the generalization ability of BP neural network and improve the accuracy of prediction model.Finally,combined with the linear processing capability of ARIMA model and the small sample and nonlinear mapping ability of SVM,and using IGS algorithm to optimize SVM parameters,the ARIMA-IGS-SVM prediction model is constructed to evaluate the effect of model in predicting R value from MAE,MSE and RMSE.Then it is found that ARIMA-IGS-SVM model has the strongest fitting ability,followed by the IGS-SVM model and the ARIMA model is the worst,which reflects the application value of the model built in this paper.To further illustrate the accuracy of the stock price range prediction model based on information granulation and BP-Bagging and ARIMA-IGS-SVM proposed in this paper,the interval prediction rate is used to evaluate the weekly forecast interval of CITIC Securities from October 23,2017 to November 17,2017.The results show that the prediction accuracy of BP-Bagging and ARIMA-IGS-SVM models for the next week is93.95% and 91.94% respectively,which is of certain reference significance for quantitative investment.