基于最大间隔最小体积超球支持向量机的多主题分类算法

摘要

针对多主题分类,结合最大间隔最小体积超球支持向量机和模糊理论,提出一种多主题最大间隔最小体积超球支持向量机来实现多主题分类.该算法首先基于最大间隔最小体积超球支持向量机,采用1-a-r方法训练子分类器,通过子分类器得到待分类样本的隶属度向量,再依据隶属度向量判定该待分类样本所属类别.实验结果表明,该算法具有较好的准确率、召回率、F1值.%For multi-subjects classification problem,a multi-subjects maximal-margin minimal-volume hypersphere support vector machine was proposed according to maximal-margin minimal-volume hypersphere support vector machine and fuzzy theory. The algorithm uses 1-a-r maximal-margin minimal-volume hypersphere support vector machine to train sub-classifiers,obtain membership vector of the sample that is classified according to the classifiers. At last it labels the subjects that the sample belongs to according to the membership vector. The experimental results show that the algorithm has higher performance on precision, recall and Fl.