A note on constrained k-means algorithms

This paper describes extensions to the k-means algorithm for clustering dat a sets. By adding suitable constraints into the mathematical program formul ation, an approach is developed, which allows the use of the ii-means parad igm to efficiently cluster data sets with the fixed number of objects in ea ch cluster. The new algorithm is presented and the effectiveness of the alg orithm is demonstrated with experimental results. (C) 2000 Pattern Recognit ion Society. Published by Elsevier Science Ltd. All rights reserved.