Balance K-Means Algorithm

K-means is the most popular clustering algorithm and many researchers pay much attention to improving it. In this paper the authors find that some features influence so much on the results of clustering. For improving the K-means algorithm, the authors design a novel balance K-means algorithm. The main idea is that we normalize all the feature values of dataset before clustering. So all the features play the same important role in the clustering, which make the k-means balanced. There are three contributions to this paper. First the disadvantages of the standard K-means are illustrated in detail. Second we design the balance K-means algorithm which all the values of features are projected into a fix range, so it can take over the disadvantage of the standard K-means and. At last the authors choose some datasets from UCI for experiments. And the results of experiments show that the balance K-means runs better than the standard K-means.