Study on the Self-Organize Selective Fusion Support Vector Machine Algorithm

Abstract

As a major statistical learning method in case of small sample, Support Vector Machine Algorithm (SVM) has some disadvantages in dealing with vast amounts of data, such as the memory overhead and slow training. we use Multi-class Support Vector Machine (MSVM) with Self-Organize Selective Fusion (SOSF) to optimize the multiple classifiers selectively, which can update the classification and self-adjust its classification performance, and eliminate some redundancy and conflicts, achieve the fusion of multiple classifiers selectively, and effectively solve the shortcoming of disturbances by the sub-samples distribution in large sample, and improve the training efficiency and classification efficiency.