Toward Ameliorating K-Means Clustering Algorithm

Abstract

Mining knowledge and predicting the behavior of the data have become major challenge with the advent of unprecedented escalation in the volume of the existing databases. Generally, clustering is adopted for voluminous and intricate data. In the present work, two techniques of K-means clustering, namely, K-means algorithm with random sampling (without realignment) and K-means algorithm with realignment sampling, are compared in terms of time taken and number of moves made for clustering the given data. The first one checks for any transfers between the clusters after inserting all the data. The second one is to check for any transfers between clusters for each new data inserted into cluster. The experimental results reveal that K-means clustering algorithm with realignment has performed reasonably well against K-means clustering algorithm without realignment.