Research And Application Based On Fuzzy Support Vector Machine

Fuzzy support vector machine(FSVM),which was set up on the basis of classical support vector machine(SVM),is a pattern recognition algorithm,this algorithm not only shares the advantages of SVM,but also can overcome the effects of noise effect on classification results.In recent years,the successful use of this algorithm makes its application scope more and more widely.Here,in view of the FSVM in the practical application,the following two aspects of improvement are put forward.On the one hand,considering that the size of data samples during training stage is so large that affect the operational efficiencies and predict time,an improved method(short for FFSVM)based on the combination of FSVM and fuzzy c-means(FCM)clustering is proposed.First of all,reeducating the dimension of the original data set by PCA,thereby clustering this data on the principal component feature sets with FCM,then extract uncertain samples by means of the role of structural uncertainty set.Finally,the FSVM is applied to uncertain samples to improve the classification accuracy.On the other hand,a novel FSVM is presented to deal with the phenomenon when the input data is fuzzy number,this method transform from the uncertain data to precise one with defuzzification function.so as to achieve accurate values of the method.The numerical experimental results show that the proposed method get lower misclassification rate compared with other algorithms.What is more,this method still can achieve better classification effect under the different parameters.