Ma JiMing
Zhengzhou University of Light Industry, 450002, Zhengzhou, China
Li XiaoJiao
Zhengzhou University of Light Industry, 450002, Zhengzhou, China
Su RiJian
Zhengzhou University of Light Industry, 450002, Zhengzhou, China
Zhang Xiang Mei
Zhengzhou University of Light Industry, 450002, Zhengzhou, China
ABSTRACT
Traditional k-means algorithm randomly select initial cluster center and the quality of clustering results depends on the selection of initial cluster center. If isolate points are selected, the algorithm iterations will increase significantly; if k points in the same class are selected, the algorithm will fall into local optimum. An adaptive method to select initial cluster center for k-means algorithm is proposed, so that initial cluster centers are located in high density area and have a certain distance with each other. Experiments show that the method has higher stability and accuracy than the traditional k-means algorithm and some other similar improved algorithms.
PDF References Citation
How to cite this article
Ma JiMing, Li XiaoJiao, Su RiJian and Zhang Xiang Mei, 2013. An Adaptive Initial Cluster Center Selection K-means Algorithm and Implementation. Information Technology Journal, 12: 5665-5668.
DOI: 10.3923/itj.2013.5665.5668
URL: https://scialert.net/abstract/?doi=itj.2013.5665.5668
DOI: 10.3923/itj.2013.5665.5668
URL: https://scialert.net/abstract/?doi=itj.2013.5665.5668