所属单位：

理学院

发表刊物：

Advances in Intelligent Systems and Computing

关键字：

中文关键字：不平衡分类；特征选择；集成学习；并行遗传算法，英文关键字：Imbalanced classification; Feature selection; Ense

摘要：

Ensemble learning currently is a popular and effective learning strategy in dealing with class-imbalance problems. For high dimensional small sample datasets, it is necessary to conduct effective feature selection to extract potentially useful information. Combined with parallel genetic feature selection algorithm with SMOTEBoost imbalanced learning algorithm, we proposed a novel imbalanced learning algorithm SMOTEBoost_PGA. It firstly utilizes PGA_GLM algorithm to select important feature, and then in each iteration it utilizes SMOTEBoost algorithm to learn the ensemble classifier on the subspace based the selected feature, and finally combines these ensemble learners. Experimental results show that the newly proposed algorithm has relatively higher AUC, F-measure, and G-mean values than many existing imbalanced learning methods. Moreover, it has approximately the shortest testing time as that of feature selection when the same number of weak classifiers are used, which is significantly faster than other imbalanced learning algorithms.

备注：

殷清燕

第一作者：

郑秀云,李体政,史加荣,殷清燕

论文类型：

期刊论文

卷号：

卷：

期号：

期：

页面范围：

页：

是否译文：

否

发表时间：

2016-12-01