The Further Research On Support Vector Machine(SVM) Algorithm

In recent years Support Vector Machine (SVM) is a research hotspot, with a solidfoundation of statistical learning theory, especially for high-dimensional, small sample,nonlinear pattern recognition problem, and can promote applied to the traditionalfunction fitting problem.This article first briefly analyze the basic theory of Support Vector Machineand Fisher Discriminant Analysis, including Support Vector Classification (SVC)algorithm, Support Vector Regression (SVR) algorithm, linear Fisher DiscriminantAnalysis (FDA) and Kernel Fisher Discriminant Analysis (KFDA) algorithm.Then, presents projection point a strategy of reducing the size of the trainingsample set for Support Vector Machine. The strategy to use the Fisher DiscriminantAnalysis method can be quickly removed a large number of non-support vector, withthe remaining samples to build the SVM training sample set. The simulation resultsshow that the algorithm can reduce the large-scale samples, and at the same timecan ensure that the classification accuracy and efficiency of the algorithm.Fitting function normally predict a relationship between variables by limitedtraining samples of the trained function, In practice due to inherent noise and isolationof training samples, the results of fitting function often do not meet the requirements byusing traditional methods. Considering the difference of feature’s correlative degree tothe regression problem becomes large, In this paper research the Support VectorRegression Machine algorithm which uses the Grey Correlation Grade as the featureweight. And extend in the application of the two-dimension Functions Fitting. Theexperiment result proves to get the better competence of regression fitting than thetraditional support vector machine.Finally, we summarize the research of this paper, and put forward somesuggestions about further study.