Research On Incremental Learning Of Support Vector Machine Based On Robustness

Support vector machine(Support Vector Machine,SVM)is proposed by Vapnik in 1990 s,the support vector machine is the first to solve the small sample data,after years of development,the support vector machine has been widely used in the regression analysis,and was used in the face recognition,handwritten numeral classification recognition.However,with the development of science and technology,the traditional support vector machine has been unable to meet the current mass data classification.In order to better apply to large-scale data classification,support vector machine incremental learning algorithm arises at the historic moment.By continuously adding new learning sample set incremental support vector machine,continuous adjustment of support vector set and the classification hyperplane can extract features better sample set,make better classification,but there is a problem in the incremental learning process,if the number of support vectors increases,with the increase of the incremental sample so,will cause the error in the classification of the classifier;if abandon part of the vector classification hyperplane has no effect,in subsequent training,this vector may become support vectors,so it will affect the classification results.In this paper,support vector machine incremental learning algorithm based on fuzzy C means clustering and central density is adopted.Fuzzy support vector machine(SVM)is used to decide support vector,and the selected support vector set is selected.The center of density ratio is used to decide the non support vector to get the selected non support vector set.The algorithm is based on fuzzy C means clustering.The original sample respectively construct support vector classification hyperplane of support vector set with incremental sample set to construct the classifier,the original sample selection support vector sets and incremental learning support vector in the sample set,the original sample of support vector set and the selection of non support vectors set with incremental sample in the selection of support vector set and the selection of non support vector set classifier,the original sample selection of support vector set and the selection of non support vectors set with new samples in the selected set of support vectors and non selected set of support vectors to construct classifier.By comparing the classification efficiency and classification effect of the above four situations,we study the influence of non support vector on support vector incremental learning robustness,and finally draw the conclusion of this paper.