An Iterated Local Search Approach for Minimum Sum-of-Squares Clustering

Abstract. Since minimum sum-of-squares clustering (MSSC) is an NPhard combinatorial optimization problem, applying techniques from global optimization appears to be promising for reliably clustering numerical data. In this paper, concepts of combinatorial heuristic optimization are considered for approaching the MSSC: An iterated local search (ILS) approach is proposed which is capable of finding (near-)optimum solutions very quickly. On gene expression data resulting from biological microarray experiments, it is shown that ILS outperforms multi–start k-means as well as three other clustering heuristics combined with k-means.

• Type: Conference
• Year: 2003
• Where: IDA
• Authors: Peter Merz
• Keywords: Combinatorial Heuristic Optimization, Global Optimization, IDA 2003, NPhard Combinatorial Optimization