Research On Stock Forecasting Method Based On Wavelet Analysis And Support Vector Machine

As the product of the market economy,the development of the stock market and the development of market economy complement each other.Stock forecasting has gradually become an important direction for domestic and foreign scholars to study and explore.In recent years,all kinds of experts and scholars on the stock market research method are no longer limited to the traditional fundamentals analysis,such as technical analysis,evolution analysis and other methods.Instead,they set more vision to the emerging mathematical methods and machine learning and other modern technology analysis methods,such as time series prediction method,neural network prediction method,fuzzy time series prediction,chaotic time series prediction method and so on.In this paper,an ARIMA-SVR stock forecasting model based on wavelet analysis is proposed.Firstly,the original data are decomposed by wavelet and wavelet reconstructed into the reconstruction part and the error part,are obtained.Reconstruction part of the main part of the data is the trend of long-term trend,the error part of the exact data and the general trend reflects the deviation.Then,the ARIMA model and the SVR model are used to predict the reconstructed and error parts respectively,and the predicted results are combined to obtain the final prediction result.This paper chooses the daily closing price of the Shanghai Pudong Development Bank from Jan.2,2014 to Aug.3,2016 as the experimental data,using the previous 600 data to predict the 30 data to explain the process in detail.And then use the forecast data of the eight stocks of Shanghai Pudong Development Bank,Guilin Tourism,Hailan House,Shenzhen Energy,Jianghe Group,Xoceco Electronics,Pathfinder and CITIC Heavy Industry to compare the ARIMA-SVR stock forecasting model based on wavelet analysis proposed in this paper with the ARIMA model and the SVR model.The simulation results show that the proposed method,the ARIMA-SVR stock forecasting model based on wavelet analysis is more accurate and is an effective stock forecasting method.