Study Of Algorithms For Support Vector Machine

Support vector machine (SVM) is a new solution to solve machine learning by optimization, which bases on statistical learning theory and can forecast intending data and fact better which is not observation. In this paper, taking out support vectors in advance and clustering method are studied by how to solve two classes support vector machine and multi-classes support vector machine.Firstly, we introduce the statistical learning theory, which is academic learning foundation of support vector machine, and the actuality,background and development of support vector machine, and discuss the usual methods solving SVM and the main ideas of two classes support vector machine.In the second chapter, we introduce the main ideas and methods of the multi-class support vector machine, bring forward two improvements for the method of half-against-half and get better training effect.In the next chapter, we introduce the main ideas of the taking-out support vectors in advance, put forward a new method of the taking-out support vectors in advance-removing non-support vectors, and find good test effect.In the fourth chapter, clustering algorithm is proposed. We introduce its background and actuality, put forward a new classification for the multi-class support vector machine, compare it with the popular algorithms, and elicit test data.In the last chapter, we summarize the paper's contents and propose some suggestions of the future work.