In this paper we propose a new approach for classification problems and apply it to eight problems. A classification problem with a large feature set is partitioned using clustering on a subset of the features. A multi-label classifier is then trained individually on each cluster, using...