Wavelet-based Research Of Chinese Stock Market Time Series

ABSTRACT:As wavelet analysis has good time frequency localization feature, since late eighties of the 20th century, it has been used widely in the fields such as signal analysis, image processing, theoretical physics, computer classification and identification, medical imaging and diagnosis, seismic exploration and data processing. In recent years, the research of analysis and prediction of actual financial data with wavelet analysis is going deeper.In this paper, we systematically conclude the basic theory of wavelet analysis and some typical models of time series analysis. We discussed the ARMA(p,q) Model which was used deeply in this paper. Especially, including the definition, recognition, parameter estimation and diagnostic test; Based on this, we introduced an ARMA(p,q) Model prediction method based on wavelet analysis, and we did empirical analysis for the Chinese Shanghai A Share Index and Chalco index by this new method, we have gained good short-period prediction result.By comparative study of data which has been processed by different wavelet functions, we found that the prediction result of data processed by Biorthogonal wavelet function is better than that of the data processed by Daubechies wavelet function. Also, the prediction method we provided has better forecast accuracy compared with the traditional method, therefore has good practical applications value. Since the data we used in empirical analysis is typical of Chinese stock market, we can see for some extent that this ARMA model prediction method based on wavelet analysis is suitable for Chinese stock market time series research.