论文部分内容阅读
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.
In order to improve the performance of algorithm classification under unbalanced datasets, an over-sampling algorithm based on negative immune is proposed, which uses negative immunization to realize the coverage of a few types of samples to generate a small number of samples that are generated manually by the center of the detector. The algorithm uses a large number of sample information to generate minority samples and avoids the lack of spatial representation of artificial samples generated by SMOTE.Through the comparison between this algorithm and SMOTE algorithm and its improved algorithm, The results show that this algorithm not only effectively improves the classification performance of a few samples, but also improves the overall classification performance.