An Analysis of Enhancement in K-Means Clustering

Abstract

Today in modern era, everyone has to retrieve the large amount of data from a vast collection of data. This process of retrieving useful data in understandable form is data mining. Big data[6] is a term for data sets that are so large or complex that old data processing applications are insufficient. Accuracy in big data might lead to more confident decision making, & better decisions could result in greater operational efficiency, cost reduction & reduced risk. Various algorithms[5] & techniques like Classification, Regression, Artificial Intelligence, Neural Networks, Association Rules, Decision Trees, Algorithm, Nearest Neighbour approach are used for knowledge discovery from databases. Clustering is an important data analytic technique which have a significant role in data mining application. Clustering is the technique of arranging a set of similar objects into a group. Partition based clustering is an important clustering technique. This technique is centroid based technique in which data points splits into k partition and each partition represents a cluster. A widely used partition based clustering algorithm is k- means clustering algorithm. But this method has problem of empty cluster. The problems could be reduced by using an enhanced algorithm. In this paper, we have analysis of the old k-means algorithm and an enhanced k-means algorithm.