The similarity of most clustering algorithms using distance measures (k-means algorithm, c-means algorithm) depend on the ability to choose the number of clusters and set of original cluster center. This greatly affects the efficiency of the algorithm implementation, even leads to poor results. In this paper, the authors build a clustering algorithm based on distance Min - Max. This algorithm is effectively guaranteed based on the nature of the Euclide distance metric (or Hamming distance) and apply the concept of closures Min - Max. In other words, the authors proposed clustering algorithm Min - Max does not depend on the choice of the earlier set of cluster center which depend only on n object clustering and determining the measure between those objects.