Genetic local repartition for solving dynamic clustering problems

Genetic local repartition is proposed to solve the dynamic clustering problem with unknown numbers of clusters. By randomly swapping and moving reference vectors between parent solutions, genetic crossovers generate six new clustering solutions. Local repartition is employed after genetic crossovers. Heuristic mean square error and minimum description length are used to evaluate the data sets with variable number of clusters. Thus, the proposed genetic algorithm automatically solves both the number of clusters and location of the clusters jointly. Test results show that the algorithm finds the clustering more efficiently than other methods considered.