Multi-class Semi-supervised Logistic I-RELIEF Feature Selection Based on Nearest Neighbor

摘要：

　　The multi-class semi-supervised logistic I-RELEIF(MSLIR)algorithm has been proposed and showed its feature selection ability using both labeled and unlabeled samples.Unfortunately,MSLIR is poor when predicting labels for unlabeled samples.To solve this issue,this paper presents a novel multi-class semi-supervised logistic I-RELEIF based on nearest neighbor(MSLIR-NN)for multi-class feature selection tasks.To generate better margin vectors for unlabeled samples,MSLIRNN uses the nearest neighbor scheme to first predict the labels of unlabeled samples and then calculates their margin vectors according to these estimated labels.Experimental results demonstrate that MSLIR-NN can improve the prediction accuracy of unlabeled data.

关键词： Logistic I-RELIEF Feature selection Multi-class classification Semi-supervised Nearest neighbor

作者: Baige Tang Li Zhang

作者单位: School of Computer Science and Technology,Joint International Research Laboratory of Machine Learnin School of Computer Science and Technology,Joint International Research Laboratory of Machine Learnin

会议类型: 国际会议

会议名称: The 23rd Pacific-Asia Conference on Knowledge Discovery and Data Mining (第23届亚太知识发现和数据挖掘国际会议(PAKDD2019)

会议地点: 澳门

会议语种:英文

页码: 281-292

在线出版日期: 2019-04-14（万方平台首次上网日期，不代表论文的发表时间）