Memetic Variable Clustering and Its Application

Table 1

Variable clustering methods for comparison.

Clustering heuristic	Initialization strategy	Stop criterion	Number of test
MCLPSO-KMEANSVAR	Initialize K centroids by MCLPSO in which each particle is initialized by uniform distribution bounded by minimum and maximum at each dimension	Maximum 2000 evaluation of KMEANSVAR	100
CLPSO-KMEANSVAR	Initialize K centroids by CLPSO in which each particle is initialized by uniform distribution bounded by minimum and maximum at each dimension	Maximum 2000 evaluation of KMEANSVAR	100
KMEANSVAR	Choose K variables randomly from the candidate variables as initial centroids	No change of the clusters	100
KMEANSVAR++	Choose K variables subsequently from the candidate variables with probability proportional to its squared distance from the point's closest existing cluster center	No change of the clusters	100