The Research About The Relationship Of Classification And Regression Of The Support Vector Machines

With the rapid development of the IT and Internet, the amount of generated date isincreasing,How to use these mass date to discover useful informations to help people to makeright decisions and guides has become a very urgent problem. As a result, machine learningtechniques come into being in such condition.Support vector machine (SVM) is developed based on statistical learning theory, it uses thestructural risk minimization principle, and through the nuclear methods to solve nonlinearproblems, This method has a good generalization ability and can solve the problem of digitsdisaster and local minimum. Support vector machine (SVM) classification algorithm based ontwo types of samples can be solved by the largest interval of the optimal classification plane,from the geometric intuition can easy to understand and the classification of the surface havea good generalization ability. From another perspective, it presents a maximum edge on thebasis of the closed convex hull contraction classification algorithm, the optimization problemof geometric meaning clear, and uses the classification algorithm in the function regression,which clearly shows the geometric meaning of regression problems.This paper mainly discusses a Nonlinear Regression Method based on One-ClassClassification after discussing Support vector machine (SVM) theory systemically. Theintroduction part explains the origin of the problem, the exploration of the topic backgroundand the present research progress.In Chapter2, statistics theory is introduced systemly,mainly introducing support vector machine (SVM) classification based on quadraticprogramming, analysising the linearly separable support vector machine (SVM) and linearinseparable. The third chapter expounds on the support vector classification problem,described in the above linear programming on the category classification and multiple classclassification algorithm a new method for nonlinear regression is proposed based onone-classification in this paper, which can uncover the relation amongone-classification,binary classification and regression. and then it gives two experimentsabout two nonlinear regressions, the results of two simulation experiments show that theproposed method is feasible and valid.finally,the summary of the work is given, and putforward some further researches.II...