不均衡数据下基于阴性免疫的过抽样新算法

Over-sampling algorithm based on negative immune in imbalanced data sets learning

为提高不均衡数据集下算法分类性能,提出一种基于阴性免疫的过抽样算法.该算法利用阴性免疫实现少数类样本空间覆盖,以生成的检测器中心为人工生成的少数类样本.由于该算法利用的是多数类样本信息生成少数类样本,避免了人工少数类过抽样技术(SMOTE)生成的人工样本缺乏空间代表性的不足.通过实验将此算法与SMOTE算法及其改进算法进行比较,结果表明,该算法不仅有效提高了少数类样本的分类性能,而且总体分类性能也有了显著提高.

To improve the classification performance of minority class,a over-sampling based on negative immune principle is proposed.In this approach,the negative immune algorithm is induced to generate a set of resprentive detectors to implement the overlapping of minority class space based on learning majority samples.The centers of resprentive detectors are regarded as the synthetic minority samples in order to resolve the imbalance problem.The majority samples are used to generate the synthetic minority samples,which can address the problem of synthetic minority oversampling technique（SMOTE）＇s lacking the ability of overlapping whole minority space using the existing minority samples.Comparing the performance of the proposed approach with SMOTE and other improved algorithms,the experimental results show that the proposed method can not only effectively improve the classification performance of minority samples,but also significantly enhance the whole classification performance.