面向不平衡数据集融合Canopy和K-means的SMOTE改进算法

针对SMOTE算法和随机森林可较好解决不平衡数据集的分类问题但对少数类样本分类效果还有待提高的问题,融合Canopy和K-means两种聚类算法,设计了C-K-SMOTE改进算法。先后利用Canopy算法进行快速近似聚类,再利用K-means算法进行精准聚类,得到精准聚类簇,最后利用SMOTE算法增加少数类样本数量,使数据趋于平衡。选取公开数据集KEEL(knowledge extraction on evolutionary learning)数据库中的不平衡数据集,结合随机森林分类模型进行了实验验证,实验表明C-K-SMOTE算法可有效平衡不平衡数据集。

The Improved SMOTE Algorithm Fusion Canopy and K-means for Imbalanced Data Sets

Aiming at the problem that SMOTE algorithm and random forest algorithm can solve the classification problem of unbalanced data set but the classification effect of minority samples needs to be improved, the improved C-K-SMOTE algorithm is designed by combining Canopy and K-means clustering algorithms. The Canopy algorithm is used to perform fast approximate clustering, and then the K-means algorithm is used for accurate clustering to obtain accurate clustering clusters. Finally, the SMOTE algorithm is used to increase the number of samples in a few classes, so that the data tends to be balanced. In this paper, the unbalanced data set in the KEEL (Knowledge Extraction on Evolutionary Learning) database is selected and verified by the random forest classification model. The experiment shows that the C-K-SMOTE algorithm can effectively balance the unbalanced data set.