Clustering searchalgorithmforthecapacitatedcenteredclusteringproblem

The capacitatedcenteredclusteringproblem(CCCP)consistsinpartitioningasetof n points into p disjoint clusters withaknowncapacity.Eachclusterisspecifiedbyacentroid.Theobjectiveistominimizethe total dissimilaritywithineachcluster,suchthatagivencapacitylimitoftheclusterisnotexceeded.This paper presentsasolutionprocedurefortheCCCP,usingthehybridmetaheuristicclusteringsearch(CS), whose mainideaistoidentifypromisingareasofthesearchspacebygeneratingsolutionsthrougha metaheuristic andclusteringthemintogroupsthatarethenfurtherexploredwithlocalsearchheuristics. Computational resultsintestproblemsoftheliteratureshowthattheCSfoundasignificantnumberof new best-knownsolutionsinreasonablecomputationaltimes.