Research On Prediction Of Stock Index Based On Combination Model Of Time Series Model And Machine Learning

The stock price index is mainly compiled by relevant financial service institutions or stock exchanges.The relative stock price statistics used to reflect the overall price level and price change trend of my country's stock market.It is an indicator that can reflect the overall price change of the stock market.Using time series analysis method to study and predict stock price indices can assist investors in formulating reasonable investment plans and thus obtain greater returns.The sequence of stock price index contains two different characteristics of linear part and non-linear part.This paper uses the combined research method of ARIMA model in traditional time series model and improved regressive SVM to study the stock price index prediction.Among the various models of time series analysis,the ARIMA model has good characteristics and is therefore widely used.But the ARIMA model cannot handle the non-linear part of the data.The support vector machine can project the nonlinear part into a new high-dimensional space and make it linearly separable,thereby directly transforming the nonlinear problem into a nonlinear separable problem.As an important index of my country's securities market,the CSI 300 Index has far-reaching research significance.However,as a kind of time series in the market economy,the sequence of the CSI 300 Index is mainly affected by various factors,and the linear and non-linear components in the data are mixed together.This article mainly adopts the design idea of the combined model,uses the ARIMA model to predict the linear part of the closing price sequence of the CSI 300 index,regards the residual as the non-linear part of the data,and uses the improved support vector regression algorithm to predict.Hybrid model,that is,first use the ARIMA model to extract the linear component of the data,and then use the improved support vector regression algorithm to process the non-linear part.After obtaining the predicted value,use the three indicators of MAE,RMSE and MAPE to compare the prediction effects of a single ARIMA model,a single improved support vector regression algorithm and a combined model,and the existing ARMA-GARCH model.The results of empirical analysis show that the prediction results of the combined model established in this paper are better than the two single models and the ARMA-GARCH model.Due to the non-tradable nature of the Shanghai and Shenzhen 300 Index and the investment and trading of a single stock under the forecast results,it still takes a lot of time and energy.Many stock index forecasting studies came to an abrupt end after obtaining the forecast results and conducting comparative analysis,lacking specific details.The formulation of investment transaction plan and the analysis of the rate of return.Therefore,the innovative point of this article is that on the basis of the prediction of the CSI 300 Index,an appropriate investment plan for the CSI 300 Index Fund is formulated according to the model prediction results of the index sequence and the index fund sequence.A comparative analysis of the rate of return of the forecasting model is carried out.The analysis shows that no matter which data is used,the investment strategy corresponding to the combination model produces a higher rate of return,and the rate of return within 10 days exceeds 3%,which means that the investment return isrelatively large.