Some Weighted Support Vector Machine And Its Applications

The support vector machine (SVM) was pointed out by Vapnik for the first time in1995. SVM has a lot of unique performance advantages in solving the small sample, nonlinear and high dimensional pattern recognition problem. Its basic principle is to find a linear classifier between two groups of samples to distinguish them while ensure the distance between two groups of samples is largest.In the first, we provide; background of our problems. Then we introduce the theoretical basis of support vector machine,including learning process consistency, boundary theory and structural risk minimization principle, etc. Then give the basic algorithm of support vector machine. Then we discussed various derivative support vector algorithm.By comparing the advantages and disadvantages of various derivative support vector machine methodswe have the theoretical preparations for the propose of new support vector machine or to improve the algorithm of support vector machine,in practical application, Some samples are more important, We hope they can be classified correctly, while Some samples are relatively less important, Therefore, We use different punishment coefficient in the optimization problem to get more accurate classification. We call this kind of support vector machine for weighted support vector machine(WSVM). And we also want to use various derivative of the advantages of support vector machine (SVM) method. So in the fourth chapter,We mainly discusses various derivatives of the support vector machine (SVM) which had weighted.One-class SVM was mainly used to slve one single classification problem, for example abnormal points search.Now we put this method applied to many classi-fication problem. By introducing the concept of ownership degrees,This method can be applied to many of the classification problems.in chapter5We are given in detail of the operating methods.The sixth chapter is algorithm verification. We adjust the number of two types of training samples. We are the first to use basic support vector machine (SVM) method to calculate. Then we give greater punishment coefficient on smaler sam-ple,and training its. Finally we found weighted support vector machine relative to the basic support vector machine still has certain advantages. The main job of this paper is to apply weighted thoughts to various derivative algorithm of support vector machine.In addition,this paper will expand the application scope of One-class SVM,we will use this method to solve binary classification problem.