The k-means algorithm is one of the most widely used methods to partition a dataset into groups of patterns. However, the k-means method converges to one of many local minima. And it is known that, the final result depends on the initial starting points (means). We introduce an efficient method to start the k-means with good starting points (means). The good initial starting points allow the k-means algorithm to converge to a “better” local minimum, also the number of iteration over the full dataset is decreased. Our experimental results show that, good initial starting points lead to improved solution.