The Application Of Wavelet Analysis In The High-frequency Financial Data Analysis

As the wavelet analysis in time domain and frequency domain has a good localized nature simultaneously, Time Domain sampling for high frequency components gradually fine step。Based on this point, In this paper, wavelet analysis method to study the high-frequency financial time series, the main ones are as follows:1. Using multi-resolution analysis theory of wavelet analysis to decompose high-frequency financial time series appropriately,making decomposed layers of data meet the balance of conditions, and fitting the every layer data by ARMA model.Using the fitting model to forecast and accumulating the forecast result, we get the actual predicted values.2. Based on wavelet variance theory,researching volatility of the month return of Shanghai and Shenzhen stock.The month volatility of this two stocks has a similar trend under the same scale,the same month and the same year.3. We estimate wavelet variance of five minutes return from shanghai and shenzhen stock by discrete wavelet tranform coefficient,and get the varivance in different scale.After logarithm tranform to the scale and varivance,we can see that there is an obvious linear relationship between them.4. Base on maximum overlap discrete wavelet transform,we can estimate the wavelet cross correlation of the five minutes return and research the relativity under the different scale and different lag.Under the small scale,the change of correlation is quick with the lag increases and the relativity of the series is little.In the big scale, the change of correlation is placid with the lag increases and the relativity of the series is big.