The Empirical Study Of Share Price Forecast Based On Support Vector Regression Theory

People have the capability to study from past data and former knowledge,andget the cognition and rules of the objective world by study and generalizing.Helpingus to forecast and judge the future phenomenon accurately.Recently,time series anal-ysis is used in financial prediction extensively,it is well used by a large number ofresearchers because of its excellent performance.Support Vector Machine based onStatistical Learning Theory is a new machine learning theory advanced by VladimirN. Vapnik in 1970s.It transform the learn problem into solving optimization methodon the basis of structural risk minimization rule,and show very good generalizationperformance.It is able to deal with nonlinear,small sample and local extremum prob-lems e?ectively.More and more researchers are absorbed in the research of SupportVector Machine,because of its superiority.In this thesis,first we introduce the Statistical Learning Theory,then we ex-pound on the theory and algorithm of the Support Vector Regression theory in de-tail.We also introduces a new method-Total Support Vector Regression Machine.Weestimate the time series model of the Shanghai composite index during January2010 and March 2011 with the two kinds of Support Vector Regression,and predictshort-term and long-term data respectively.The empirical results show that Sup-port Vector Regression in prediction shows strong forecasting ability.And this TotalSupport Vector Regression Machine has better forecasting ability than the generalSupport Vector Regression.Then we forecasted the stock bull market, bear mar-ket,dull and overall market by Total Support Vector Regression Machine.Finally weput the thought of feedback from neural network applied to the Support VectorRegression, made the prediction result is more accurate.