The k-means method has been shown to be effective in producing good clustering results for many practical applications.
However, a direct algorithm of k-means method requires time proportional to the product of number of patterns and number of clusters per iteration. This is computationally very expensive especially for large datasets.