Multi-label feature selection based on multi-granulation separability

Erliang Yao, Deyu Li, Yuhua Qian, Xiaozhen Fu

Abstract: Highlights•We propose a multi-granularity-based approach for multi-label feature selection.•A multi-level granulation algorithm is designed to mine data granularity structure.•A multi-granulation separability score is defined to evaluate feature importance.•Extensive experiments demonstrate the advantages of multi-granulation separability.