The Research of Simplification of Structure of Multi-class Classifier of Support Vector Machine

Because of the unique property in pattern recognition and in regression analysis, support vector machine(SVM) has become the topic of research recently. The advantages of SVM mainly lie in its capabilities of processing non-linear and highly dimensional data problems. Unextended SVM is very suitable for solving two-class classification problems. For multi-class classification, however, it should be converted into many of two-class classification problems, and can be constructed many of two-class classifier correspondingly. But the case results in more complexity of classifier structure, and so leads to decrease of decision speed. In order to get a fast classification, a new multi-class classifier with simplified structure is put forward so that the number of subclassifiers and decisive time are reduced greatly. The accuracy and complexity are also contrastively analyzed here. The validity of the new classifier is proved by simulated experiments.