基于粗糙集的多标记专属特征学习算法

Multi-label Learning with Label-specific Features Based on Rough Sets

基于专属特征的多标记学习算法使用K-Means聚类算法对标记的正反样例进行聚类,进而构造每个标记的专属特征.但该方法对标记和专属特征之间的相关性缺乏理论性地探究,而且K-Means聚类方法仅仅局限于数值属性数据聚类.对此,一个基于粗糙集的多标记专属特征学习算法(R-LIFT Algorithm)被提出,其使用粗糙集的约简算法来计算每个标记的专属特征.该算法选取的专属特征是原始特征,具有直观意义,并且能够从理论上保证专属特征与标记之间具有较强的相关性.实验表明,R-LIFT算法能够有效地学习专属特征,并进一步提高多标记学习算法的性能.

Multi-label learning algorithm with label-specific features performs clustering analysis on positive and negative instances of each label, and then constructs label-specific features. However,the algorithm cannot guarantee in theory the relevance between label- specific features and labels. And K-Means clustering method is only limited to the numeric data. In this paper, a multi-label learning al- gorithm with label-specific features based on rough sets named R-LIPT Algorithm is proposed, which uses the attribute reduct algo- rithm in rough set theory to compute specific features of each label. The label-specific features constructed by R-LIPT algorithm are a part of original features, which have intuitive significance and easy to be understood. Moreover,R-LIFT algorithm can theoretically en- sure a strong correlation between the label-specific features and labels. Experiments on real-life data sets show that the proposed ap- proach can effective find the label-specific features and,moreover,improve the performance of multi-label learning.