Application of Nonstationary Time Series Prediction to Shanghai Stock Index Based on SVM

Abstract

With the development of computer software and hardware system, machine learning methods are more and more used in various industries of social development. In the aspect of stock index prediction, the current prediction method has gradually changed from the traditional statistical analysis method to the artificial intelligence analysis method. Based on the original sample data, this paper uses support vector machine regression (SVR) model to predict the opening price of Shanghai stock index. The parameters of SVR model are optimized and debugged by grid search method (grid), particle swarm optimization (PSO) and genetic algorithm (GA). The analysis results show that the three types of support vector machine prediction models based on the original sample data can fully reflect the time-varying law of stock index and have high prediction accuracy. Among them, genetic algorithm support vector machine regression (GA-SVR) model shows that the minimum root mean square error (RMSE) is 14.730 and the minimum average absolute percentage error (MAPE) is 0.375%. GA-SVR model has good prediction effect and has certain significance for the prediction of stock price.