Multi-SVM Fuzzy Classification and Fusion Method and Applications in Bioinformatics

Support vector machines (SVMs) have been widely used for data classification. Especially, SVMs are suitable for classifying biological and biomedical data, which characteristically have a few data samples with relatively large number of input features. In many cases, SVMs show good generalization performance. Still in many other cases, the generalization ability needs to be improved. In particular, when an user develops one SVM model, it may be difficult to obtain a satisfactory generalization result. In this paper, a multi-SVM fuzzy classification and fusion model (MSFCF) has been proposed to combine multiple SVMs to enhance the generalization ability of SVMs. The approach constructs a fuzzy fusion system to combine output values from several SVMs in respect to the accuracy information of each SVM. The final decision is determined based on all SVMs. The experimental results show that the proposed MSFCF has more stable and robust generalization ability than a single SVM.