Research And Application Of SVM Based On The Mixed-kernel Function

Support vector machine (SVM) is a classific method based on statistical learning theory, it is put forward by Vapnik, Make the lower dimensional feature space through a nonlinear transformation to get a high dimensional feature space, then looking for the optimal separating hyperplane in the new space, which it will be transformed separable problem from the inseparable problem. SVM show strong advantages in dealing with small samples, high dimension, nonlinear in practical problems, so,it is a very important method of machine learning.Kernel function is the core of the support vector machine (SVM), different kernel functions directly affect the performance of support vector machine, kernel function performance become one of core issues in the study of support vector machine (SVM).Firstly this article introduce the theory of support vector machine (SVM) and kernel function, based on the research of the nature of SVM and kernel function and kernel function test of four common (linear kernel function and polynomial kernel function, gaussian kernel function, sigmoid kernel function) performance on different data sets, select the two or more optimal kernel function to construct new hybrid kernel function, and use genetic algorithm to learning the weight coefficient and kernel parameters in the mixed kernel function by automatic optimization, and test the performance of the mixed kernel function on different data sets, and carries on the analysis comparison with the single core test data results, come to a conclusion. verify the mixed kernel function of support vector machine is practical. Finally, the research work is summarized, and Put forward the further research problems.