Comparing initialisation methods for the Heuristic Memetic Clustering Algorithm

In this study we investigate the effect five initialisation methods from literature have on the performance of the Heuristic Memetic Clustering Algorithm (HMCA). The evaluation is based on an extensive experimental comparison on three benchmark datasets between HMCA and the commonly-used k-Medoids algorithm. Analysis of the experimental effectiveness and efficiency metrics confirms that the HMCA substantially outperforms k-Medoids, with the HMCA capable of finding bestter clusterings using substantially less computation effort. The Sample and Cluster initialisation methods were found to be the most suitable for the HMCA, with the results of the k-Medoids suggesting this to be the case for other algorithms as well.