基于模糊支持向量机的多标签分类方法改进

摘要

Multi labels sorting of one-against-all support vector machine (SVM)exists the problems that the sample is sorted into training set while undefined area can′t be acquired,also,label without clear deci-sion function has vague area.So,a multi labels sorting improved method (FSVMi)was put forward based on vague SVM.By merging multiterm decision boundary and allocating corresponded subordinate function for each label,it is verified by experiment that it has superiority over existed method.%One-against-all支持向量机的多标签分类存在将样本分类到训练集无法获取标签的"未定义"区域和没有明确决策函数的标签模糊区域的问题.对此提出一种基于模糊支持向量机的多标签分类改进方法(FSVMi).该方法通过将多条决策边界合并,并为每个标签类分配相应的隶属函数.实验结果表明,相比于现有方法,该方法更具有优越性.