Fuzzy information gain ratio-based multi-label feature selection with label correlation

Multi-label feature selection aims to mitigate the curse of dimensionality in multi-label data by selecting a smaller subset of features from the original set for classification. Existing multi-label feature selection algorithms frequently neglect the inherent uncertainty in multi-label data and fail to adequately consider the relationships between features and labels when assessing the importance of features. In response to this challenge, a Fuzzy Information Gain Ratio-based multi-label feature selection considering Label Correlation (FIGR_LC) algorithm is proposed. FIGR_LC evaluates feature importance by combining the relationship between features and individual labels, as well as the correlation between features and label sets. Subsequently, a feature ranking is established based on these feature weights. Experimental results substantiate the effectiveness of FIGR_LC, showcasing its superiority over several established feature selection methods.