Research On Stock Index Prediction Based On ARFIMA-GRACH-SVM Model

Since Peters put forward the fractal market hypothesis in 1990,the research on the nonlinearity of financial time series has never stopped,including the research on long memory and conditional heteroscedasticity.The nonlinear model obtained in the study enables scholars to solve some problems that cannot be solved by traditional time series methods.In addition,with the continuous development of computer,a large number of scholars apply machine learning to the prediction of financial data.In 2020,the world financial market was severely affected by COVID-19.As the world’s financial center,the US stock market triggered four circuit breakers in March.Series of shocks also had a certain impact on China’s stock market.The stock market is a barometer of a country’s economy,so it is very important for ordinary investors and national regulators to correctly judge the trend of the stock market.In this paper,empirical mode decomposition algorithm is used to decompose the data into fluctuation part and trend part,and nonlinear time series model and machine learning model with external variables are used to establish the estimation model,in order to achieve better prediction results.This paper can be divided into seven parts.The first chapter mainly introduces the development and application of each model in finance;the second,third,fourth and fifth chapters introduce the theoretical knowledge,model methods,modeling steps and optimization criteria needed in this paper in detail;the sixth chapter is the empirical part of this paper,which establishes linear combination forecasting model and SVM combination forecasting model based on arfima-garch-svm for CSI 300 and CSI 500 respectively The seventh chapter is the summary and Prospect of the model.The innovation of this paper is to decompose the data by EMD method,and combine with ARFIMA model,in order to reduce the difference order,filter the noise of the data,make full use of the information in the data,and then use support vector machine to estimate the fluctuation direction of GARCH model,so as to reduce the randomness of the fluctuation of traditional arfima-garch model In addition,the combination method of this paper uses machine learning method to deal with the random number that the GARCH model prediction value needs to be multiplied by in a nonlinear way,in order to reduce the error.Secondly,for the problem of multi-step prediction using external variables in SVM model,this paper uses the establishment of multiple models to deal with.Finally,from the comparison results of various models,it is found that when the data performance is strong nonlinear,the combination model using SVM can correct the prediction results of the previous parts,and the prediction results are signifi-cantly better than the combination model only using linear combination.Finally,this paper selects the combination model to make short-term prediction of the stock index during the epidemic period,and uses the combination model and GARCH model prediction results as the investment basis to provide feasible suggestions for investors:the combination model established in this paper is basically reliable,in-vestors can decide the next week’s investment choice according to the prediction value of the combination model in the next five days It is necessary to be cautious in case of large fluctuation,which may lead to the failure of the combined model.