Volume-based clustering for arbitrary shaped clusters. (1st January 2013)

• Record Type: Journal Article
• Authors: Bishnu, Partha Sarathi; Prasad, Saurabh; Bhattacherjee, Vandana
• Is Part Of: International journal of computational vision and robotics. Volume 3:Number 3(2013)
• Page Start: 167
• Page End: 181
• Publication Date: 2013-01-01
• Subjects: data mining -- clustering -- noise; Computer vision -- Periodicals; Robotics -- Periodicals; Artificial intelligence -- Periodicals; 006.3705
• Languages: English
• ISSNs: 1752-9131

Abstract

Volume-based clustering is a clustering technique for identifying arbitrary shaped clusters. The main aim of this paper is generation of arbitrary shaped clusters by forming sub-clusters which are merged to identify the actual shape of the clusters and two phase outlier detection and removal is conducted. By varying a user defined parameter e, user can get the desired number of clusters. Experiments were conducted on different sets of real and synthetic datasets to test our proposed algorithm and results are compared with other existing algorithms.