The Research On Data Mining SVM Method

Support vector machine is one of many data mining algorithms, it is a tool to solve the classification s problems by optimization. Currently, support vector machine has been succes-sfully applied in many fields, such as health care, finance, agriculture, education and so on.And it has become a hot topic of concern. Both in theory and algorithms, researchers have achieved satisfactory results.But when the data size is large, the traditional method in training process require too much storage space, take too long time and not easy to apply. Directed to the above problems,after in-depth study of support vector machine theory, we have proposed the corresponding improvement method, details are as follows:Firstly, for support vector classification, we have used the method of minimum distance to pre-drawn boundary vectors, and then used support vector classification machine adjustable entropy function method to train pre-drawn boundary vectors. At last, we have used three experiments verify the feasibility of the method, and we have analyzed some advantages of this method in terms of storage and training time.Secondly, for support vector regression, we have used the method of selecting a class ce-nter contains all support vectors as the border vector set, then the set has been trained with support vector regression adjustable entropy function. Through experimental test the classific-ation accuracy and performance advantages of the method.