An Improved K-Means Algorithm Based on Multiple Feature Points

Abstract

K-means algorithm is one of the most popular clustering algorithms. However, it is sensitive to initialized partition and the circular dataset. To attack this problem, this paper introduced an improved k-means algorithm based on multiple feature points. The algorithm selects a number of feature points as cluster centroids unlike the traditional algorithm which only uses one centroid. In addition, the algorithm calculates the weighted distance to distribute the data point and to build the new feature points set. Theoretical analysis shows that the improved algorithm and the traditional algorithm are in the same order of magnitude. The experimental results on artificial data show that the improved algorithm is better than the traditional algorithm. Experiment on real data gives appropriate parameters.