Applications Of HILBERT-HUANG Transform On Financial High Frequency Time Series

With the rapid development of computer science and information technology, investors can fetch a variety of transaction data from securities market more and more easily and timely. With a further research on each type of transaction data, data by minutes, even by tick, gradually attracted the attention of investors. The high-frequency financial time series are abstract form these data. People apply each type of data analysis method on the highfrequency financial time series, attempt to reveal the inherent characteristics and forecast the price trend.By assuming a linear relationship between each variable, the classical time series analysis methods attempt to find out the relationship between independent variables and dependent variable so as to predict the value of dependent variable. Another method is analyzing the time or frequency characteristics and extension the endpoint to forecast the value of next time. These traditional analysis methods, such as Fourier Transform, usually require the time series stationary. But in real life, time series formed by the price of portfolio and stock index features, is often nonlinear and nonstationary. Therefore the traditional data analysis methods cannot work well on data mining of financial time series.With the development of modern signal processing technology, HilbertHuang Transform is applied more and more in nonlinear and non-stationary data and has achieved good results in many application fields. The main content of this paper is the analysis and prediction of high frequency financial time series by Hilbert-Huang transform. Based on the Shanghai and Shenzhen 300 stock index features minute data, the empirical mode deposition method of Hilbert-Huang Transform is applied on the highfrequency time series. And Elman neutral network is used to forecast the next value of each intrinsic mode function. The test has been done based on the sample data. Based on the predict model, a trading strategy is formed and tested by using Shanghai and Shenzhen 300 stock index features price data. The result of the test shows that the forecast module based on Hilbert-Huang Transform and Elman neutral network can better describe the character of price data of Shanghai and Shenzhen 300 stock index futures and predict for the non-stationary time series.