Font Size: a A A

Application Of Wavelet Neural Networks In Financial Time Series

Posted on:2008-12-05Degree:MasterType:Thesis
Country:ChinaCandidate:Y LuoFull Text:PDF
GTID:2189360215970624Subject:Computational Mathematics
Abstract/Summary:PDF Full Text Request
This dissertation mainly studies the application of wavelet neural networks in financial time series. After the revision of Efficient Market Theory(EMT), we establish the Fractal Market Theory(FMT) framework, and go on with the research on several issues based on the fractal characteristics of financial markets: wavelet transform to de-noise financial time series; nonlinear cointegration modeling; capital assets pricing in fractal markets. The main work is as the following:First, we analyse the defect of traditional method and present wavelet transform theory to de-noise origin series. Some problem about how to determine several key parameters such as wavelet function, threshold rule and decomposition level are discussed.Second, the paper establishes a wavelet neural network model and presents its properties. Wavelet neural network is introduced into the nonlinear cointegration modeling, and the modeling method is presented. Shanghai and Shenzhen stock markets show that the performance of wavelet neural is more satisfactory than that of BP neural network in the estimation of nonlinear cointegration function. The experiments results also indicate that there's nonlinear cointegrated relationship between two markets.Finally, the dissertation gives out the assets pricing theory in fractal markets by utilizing nonlinear cointegration theory, and sets up wavelet neural network models for capital assets pricing in fractal markets. The experiment of Shanghai stock market shows that the pricing method we present above is superior to Capital Asset Pricing Model.
Keywords/Search Tags:wavelet transform, wavelet neural network, financial time series, de-noising, nonlinear cointegration modeling
PDF Full Text Request
Related items