The traditional _means randomly selects initial cluster centers and divides the sample points by Euclidean distance in the original space

so the accuracy classification isntt enough. The min_max algorithm obtatins more improvement than the traditional algorithm.However

the traditional algorithm and the min_max algorithm neglect the irrationality of the classification by Euclidean distance.The improved algorithm maps the input vectors into the feature space and determines the initial centers

at last clusters by the distance between two points in the feature space. The improved algorithm is evaluated on available datasets called iris and wine.