A Comparative Approach to Evaluate Different CVIs using Grid K-Means and Improved K-Means Clustering

Abstract

Talk about the era we are living in today, is not limited to just include the living creatures but also the most important aspect upon which these living creatures are now more rely i.e., “Information”. There is an abundance of information against every aspect available on this planet. So, the term “Data” is too concise to encapsulate this “Information”. Hence, the era of Big Data come into existence. Data is now big enough (in terms of volume, variety, value, veracity, velocity) that without proper techniques and methods it is not possible to frame a definite set of knowledgeable data. A need to mine this deep ocean of information to get knowledgeable data results in the various Data Mining techniques. In this paper, a reflection of all the major data analysis techniques and how the traditional models are replaced by the new emerging technologies based on machine learning or deep learning is presented. Partition-based clustering is the most commonly used technique of unsupervised learning; in this paper, Improved K-Means and Grid K-Means algorithms are used to form clusters for four publically available datasets. The effectiveness of these clusters is evaluated using seven different Cluster Validity Indexes. Results show that VCVI-index and BVCI-index outperform among all other CVIs. © 2020, World Academy of Research in Science and Engineering. All rights reserved.