Membership based on combining cluster center with affinity in FSVR and its application in soft sensor modeling

Support vector machine (SVM) is an effective method for resolving regression problem. However, tradition SVM is very sensitive to noises in the training sample. In order to overcome this problem, fuzzy support vector regression (FSVR) based on combining cluster center with affinity is proposed in this paper. The fuzzy membership is defined not only by the distance between a point and its cluster center, but also by the two different points of the sample, which is depicted as the affinity between them. And the method of soft sensor modeling based on FSVR with the new membership function is proposed. Simulation results for artificial data show the proposed method gives good performance on reducing the effects of noise and improves the regression accuracy and generalization.