Artificial Bee Colony for Multi-Label Feature Selection: A Logic-Guided Approach With Correlation-Driven Refinement

Emrah Hancer, Bing Xue, Mengjie Zhang

Abstract: Multi-label feature selection is an emerging topic that focuses on selecting the relevant and informative features (attributes or variables) from a dataset when dealing with multi-label classification (MLC) problems. In the context of MLC, where feature interactions can vary significantly, evolutionary computation (EC) techniques might provide a more adaptive and flexible feature selection mechanism. Despite this potential, there is a noticeable scarcity of applications employing EC in the existing literature. In this paper, we introduce an ABC variant for multi-label feature selection (called LNABC_ES) that incorporates logic operators and a natural survivor selection, along with a filter-based elimination strategies. The LNABC_ES approach begins by narrowing down the potential feature pool using a filter-based pre-elimination strategy. The LNABC_ES algorithm then searches for an optimal feature subset within this refined space. Experimental results underscore the superior performance of LNABC_ES in comparison to various EC and traditional approaches. To the best of our knowledge, this study represents the first application of the ABC algorithm in the domain of multi-label feature selection, marking a significant advancement in the field.

External IDs:dblp:journals/tetci/HancerXZ25a