Title: Enhancing clustering accuracy by finding initial centroid using k-minimum-average-maximum method
Authors: S. Dhanabal; S. Chandramathi
Addresses: CSE Department, Jansons Institute of Technology, Coimbatore, Tamilnadu, India ' Sri Krishna College of Technology, Coimbatore, Tamilnadu, India
Abstract: Determining 'the initial seed for clustering' is an issue in k-means which has attracted considerable interest, especially in recent years. Despite its popularity among clustering algorithms, k-means still has many problems such as converging to the local optimum solutions, the results obtained are strongly depends upon the selection of initial seeds, number of clusters need to be known in advance etc. Various initialisation methods were proposed to improve the performance of k-means algorithm. In this paper, a novel approach, k-minimum-average-maximum (k-MAM), is proposed for finding the initial centroids by considering distance on extreme ends. The proposed algorithm is tested with UCI repository datasets and data collected from Facebook. We compared our proposed method with simple k-means and k-means++ initialisation method based on efficiency and effectiveness. The results show that the proposed algorithm converges very fast with better accuracy.
Keywords: clustering; k-means; initialisation techniques; k-minimum-average-maximum; k-MAM.
International Journal of Information and Communication Technology, 2017 Vol.11 No.2, pp.260 - 274
Available online: 04 Sep 2017Full-text access for editors Access for subscribers Purchase this article Comment on this article