The Application Of Wavelet Analysis For Denoise In Trend Trade

Beginning with noise trade, this thesis introduced three classic theorieswhich doubted the Efficient Markets Hypothesis. These theories are FinancialMarket Microstructure Theory, Behavioral Finance, Fractal MarketHypothesis.Based on above discussion, this thesis believed that the market was nottotally efficient, thus trend trading can exist in the market for a long period.However, trend trading still has its weakness which may result in a largereturn retracement in a vibrant market and suffer a liquidation risk.In order to relieve the retracement, this thesis tries to filter any possibleshort term trading noise by the method of wavelet analysis. In this way, themarket trend gets highlight, the trend trading has less interference from noisetrading, so the trading performance will be improved.To achieve this purpose, this thesis refers to relative wavelet analysistheories, and employs HAAR wavelet to analyze five kinds of market datawhich are HS300Index daily, HS300Index15minus, Shanghai CompositeIndex daily, Shenzhen Composite Index daily and decompose them intoone-level and two-level, then de-noise wavelet coefficients in each levelswith different algorithms which are default threshold wavelet de-noisemethod, given threshold wavelet de-noise method and compulsory wavelet de-noise method. Finally, this thesis gets several de-noised market data.By comparing the history performance on the raw market data andrespective de-noised market data with same trend trading strategy, the role ofwavelet de-noise in improving the performance of trend trading can beverified. The results showed that wavelet de-noise played an important role inimproving the performance of trend trading.This thesis also discovered that two-level decomposition outperformedone-level decomposition and the simplest compulsory wavelet de-noisemethod was not worse than others.So this thesis believes wavelet de-noise technology has a substantialapplication value in the trend trading.All of the wavelet transformation and de-noising involved in this thesisare accomplished in MatLab and trend trading strategy simulations areaccomplished in MultiCharts. All codes are presented in appendix.