Research And Application Of Support Vector Machine

Support Vector Machine(SVM)is based on Statistical Learning Theory (SLT)'s Vapnik-Chervonenkis(VC)Dimension and Structure Risk Minimizing (SRM).SVM owns the characters of optimal,kernel,and high generalization. Because of its powerful learning ability,SVM is more and more attached importance by both native and abroad scholars.Macroeconomy is made up of lots of economic factors which affect each other. Our nation's macroeconomy is reforming.Because of the economic system and market rulers are being constructed,to predict the macroeconomy is more difficult and more valuable.On the background,the dissertation studies on SVM and applies SVM to predicting macroeconomy.The main results of the dissertation are as follows:1.To expatiate the basic theory of SVM,to study and analyze the limitation of Empirical Risk Minimizing(ERM)and advantage of SRM,and to make a conclusion of SVM study and application.2.The study of SVM classify algorithm and regression algorithm.To study how to deal with missing data and how the method of validation works.3.The study of Group Method Data Handling(GMDH).An approach based on Objective System Analysis(OSA)and SVM is proposed for feature extracting.4.To study the methods of predicting macroeconomy.Some disadvantage of those methods is found.Economic data owns the characters of high dimensions, small samples,time series,multi-subject and multi-layer.The approach that applies SVM to economy is proposed.5.In the need of a practical project,a macroeconomic data warehouse has been built.SVM is applied to macroeconomy.The experimentations manifest that applying SVM classify and regression to economic index such as GDP has a good result.