Font Size: a A A

Application Of Non-Stationary Time Series To CSI 300 Index Based On Wavelet Transform And Support Vector Machine

Posted on:2018-01-31Degree:MasterType:Thesis
Country:ChinaCandidate:J C ChenFull Text:PDF
GTID:2428330620957837Subject:Statistics
Abstract/Summary:PDF Full Text Request
Stock index forecast is a nonlinear complex system with multi-factor influence and exponential dynamic fluctuation.The accurate forecast of stock index is able to provide reference for securities investment decision.Along with the development of computer hardware and software system,machine learning methods are more and more applied to various industries of social development.Nowadays,the stock index forecasting method has gradually changed from the traditional statistical analysis method to the artificial intelligence analysis method.In this paper,wavelet transform is used to reconstruct the opening price of the CSI 300 index to replace the opening price in the original sample data,and then constructs two different samples.The parameters(C,?)of SVR models were selected by three different methods of grid search(GRID),particle swarm optimization(PSO)and genetic algorithm(GA).The optimized parameters were used to predict the opening price of the test samples.The predictive results shown that the SVR model with GRID(GRID-SVR),the SVR model with PSO(PSO-SVR)and the SVR model with GA(GA-SVR)models were capable to fully demonstrate the time-dependent trend of stock index and have significant prediction accuracy.The minimum root mean square error(RMSE)of the GA-SVR model was 14.730,he minimum mean absolute percentage error(MAPE)equaled to 0.375%.According to the trend of the CSI 300 index,the modeling data is divided into the stable market,the volatile market and the fluctuating market.Different market data were established to the SVR models,and by comparing the predictive result of these models to select the optimized models.In the analysis of the stable market,the grid search method-support vector machine regression(GRID-SVR)model has the highest prediction accuracy,the RMSE is 5.650 and the MAPE is 0.197%.The results of the volatile market show that the WT-GRID-SVR model is the best predictive model.(RMSE,MAPE)values were equal to(20.727,0.435%).WT-PSO-SVR model exhibits the best predictive performance in the prediction of the fluctuating market.The value of(RMSE,MAPE)were(13.148,0.322%).The empirical results show that the SVR models effectively predict the opening price of the CSI 300 index.In the volatile markets and the fluctuating markets,the sample data by wavelet decomposition and reconstruction could to improve the prediction accuracy of SVR models.
Keywords/Search Tags:CSI 300 Index, Support Vector Regression, Grid Search, Particle Swarm Optimization, Genetic Algorithm
PDF Full Text Request
Related items