The Research On Forecasting Of Financial Time Series Baesd On Wavelet Analysis And Neural Network

Wavelet analysis has been greatly valued by numerous scientists and engineers for its Time-Frequency localized features and been widely applied in image processing, identification of pattern, Geological Exploration, Medical Imaging diagnosis and Numerical Calculation. It has begun to be used in the field of economy and finance to deal with the data of time series.Financial time series is a primary data type in the application of financial area. Analyzing predicting and controlling of such kind of data is the basic work of the economic and financial activity. Financial time series is composed of bond profit, foreign exchange rate, stock price, futures priceless.In this paper, financial time series is chosen to be studied by using the model that combines Wavelet theory and Neural Network. Then we use the Shanghai Composite Index to train the model and forecast the return of index.Firstly the primary developing procedure of financial time series analysis method is introduced, and explains why this research topic is chosen. We divide the research method of time series into two categories: Econometrics modeling, Data Mining. Then the advantage and disadvantage of each method are discussed. After comparing we focus on the data mining method and give a detailed illustration of Neural NetworkThen the wavelet analysis theory is discussed. The definitions of Continuous Wavelet Transform (CWT), Discrete Wavelet Transform (DWT), and Multi-Resolution Analysis (MRA) are offered. After this the characters of different wavelet functions are discussed. At last two kinds of algorithms for wavelet transform (mallat, Atrous) are compared.At last we use the Shanghai Composite Index to train and test the model. The index is decomposed into several levels using wavelet transform then the decomposed series are trained and forecasted independently. At last we use wavelet reconstruction to build the forecast of the original return series. By comparing with the ordinary BP NN we find the model introduced in this paper achieved good results.