计算机应用研究2018，Vol.35Issue(4)：1023-1027,5.DOI:10.3969/j.issn.1001-3695.2018.04.014
面向不平衡数据分类的复合SVM算法研究
Hybrid SVM algorithm oriented to classifying imbalanced datasets
摘要
Abstract
In order to improve the classification accuracy of traditional support vector machine (SVM) for imbalanced datasets,solving the problem that classifier had a low performance on minority class,this paper proposed a hybrid SVM algorithm.It combined adaptive synthetic sampling(ADASYN) algorithm with different error cost(DEC) algorithm to improve the bias of hyperplane caused by imbalanced datasets,and then it introduced a new correction algorithm to prediction model so as to improve the prediction model's adaptability to different data characteristics.It tested the proposed algorithm on 7 sets of realworld imbalanced datasets from UCI database.The experiment result shows that the hybrid SVM algorithm is able to surpass or match the state-of-the-art algorithms on each dataset,and it increases the classification performance by an average of 2.0％ to 20.9％.It shows that the proposed algorithm is effective and robust.关键词
不平衡数据/支持向量机/自适应合成采样/不同错误代价/修正算法Key words
imbalanced datasets/support vector machine/ADASYN/DEC/correction algorithm分类
信息技术与安全科学引用本文复制引用
刘东启,陈志坚,徐银,李飞腾..面向不平衡数据分类的复合SVM算法研究[J].计算机应用研究,2018,35(4):1023-1027,5.基金项目
复旦大学国家重点实验室开放基金资助项目(2015KF009) （2015KF009）
中央高校基础科研计划资助项目(2015QNA4018) （2015QNA4018）
