where zi represents ith pattern belongs into cluster Cj. Here, the goal is to obtain a partitioning of the data set, such that E is minimized. The distortion criterion is valid for cluster sample dense as well as the small differences in the number of various clustering samples. The procedure for the proposed algorithm can be summarized as follows: Setp 1: Randomly Initialize the positions of food sources (each food source being a set of centroids) and use the k-means algorithm to finish clustering task for all produced positions and compute the fitness value of each group of centroids using (6).