Submitted to Pattern Recognition Letters

In multi-label learning, each instance in the training set is associated with a set of labels, and the task is to output a label set whose size is unknown a priori for each unseen instance. Common approaches to multi-label classification learn independent classifiers for each category, and perform ranking or thresholding schemes in order to obtain multi-label classification. In this paper, we describe an original method for multi-label classification problems derived from a Bayesian version of the K-nearest neighbor (KNN), and taking into account the dependencies between labels. Experiments on simulated data sets and benchmark data sets show the usefulness and the efficiency of the proposed method compared to other existing methods. I