Decomposition and Forecast for Financial Time Series with High-frequency Based on Empirical Mode Decomposition

Abstract In this paper, the empirical mode decomposition (EMD) of the wavelet transformation is introduced into the processing of financial time series with high frequency. The high-frequency data are decomposed with EMD at first. Then the evolutionary law and development trend of each component of intrinsic mode function (IMF) are explored in different time scales. Finally, forecast model are reconstructed by using the IMF components. Using this forecast model, the time series of oil futures at 5 minute intervals as samples is analyzed. The results showed that there are quasi-cycles with 10, 27, 80, 150, 370, 860, 1290 points of data in the financial time series. Furthermore, we can make more accurate forecast with EMD by extracting the IMF components with the different volatility cycle in thefinancial time series.