| Stock market changes so fast ,the author tries to find out the variety rule of stock index, a time series which varies with the time, so that we can forecast stock index , guide effectively stock investment. Stock index is sensitive of many factors. Among them , some are from in house and other outside. They have nonlinear and complex coupled relation , so that the forecast of stock index has to think over more factors than other prediction systems.This paper brings forward the wavelet—support vector machine model , and pick up Symlet wavelet function to do discrete wavelet transform to Shanghai stock aggregative index number. Then the author utilizes multiresolution character to decompose time series into approximation and details(this paper decomposes time series into tri-level).By way of effective forecasting, to do phase space reconstruct to forecasting series decomposed ,namely transform one dimension time series into matrix to get the correlation between data and dig up more information as possible . After transformation, we get new learning samples.Then we use SVM to train new samples to get regression model. In the end ,we deal with data and reconstruct them to get final forecasting data of time series.Among tendency research of stock variation nonlinear method—which is typical of chaos, neural network, wavelet analysis and support vector machine—has rapidly grown up for twenties years and has achieved some pioneering returns.Wavelet analysis method, nonlinear chaos theory and support vector machine theory rapidly grow up among nonlinear time-series analysis by right of excellent theory to offer a new way of stock technical analysis. This paper use above-mentioned theory to set up the forecast model of Shanghai stock aggregative index number, then compare between the forecast data and actual data, and show the good result. To analyze the final result, find the creator , and point out the advantages and the defects . |
PDF Full Text Request