Exploiting label dependency and feature similarity for multi-label classification

Multi-label classification is an emerging research area in which an object may belong to more than one class simultaneously. Existing methods either consider feature similarity or label similarity for label set prediction. We propose a strategy to combine both k-Nearest Neighbor (kNN) algorithm and multiple regression in an efficient way for multi-label classification. kNN works well in feature space and multiple regression works well for preserving label dependent information with generated models for labels. Our classifier incorporates feature similarity in the feature space and label dependency in the label space for prediction. It has a wide range of applications in various domains such as in information retrieval, query categorization, medical diagnosis and marketing.