Online Multi-label Group Feature Selection

Abstract

Feature selection for multi-label learning has received intensive interest in recent years. However, traditional multi-label feature selection are incapable of considering intrinsic group structures of features and handling streaming features simultaneously. To solve this problem, we develop an algorithm called Online Multi-label Group Feature Selection (OMGFS). Our proposed method consists of two-phase: online group selection and online inter-group selection. In the group selection, we design a criterion to select feature groups which is important to label set. In the inter-group selection, we consider feature interaction and feature redundancy to select an optimal feature subset. This two-phase procedure continues until there are no more features arriving. An empirical study using a series of benchmark data sets demonstrates that the proposed method outperforms other state-of-the-art multi-label feature selection methods.