An Ameliorated Partitioning Clustering Algorithm for Large Data Sets

Abstract: K-medoid clustering algorithm is broadly used for various practical applications. Original K-medoid algorithm use to take initial centroids and medoids arbitrarily that bears on the resulting clusters and it leads to unstable and empty clusters which are meaningless and also amount of iterations can be rather high so K-Medoid is not a substitute for big databases because of its computational complexity. Also the original k-means algorithm is computationally costlier and involves time relative to the product of the number of data items, number of clusters and the number of iterations, the time complexity of K-means is O (tkn) where t is the amount of iterations. Though K-means algorithm usually leads to better outcome, it does not scale well and is not time efficient. Ameliorated k-Medoid clustering algorithm will have the accuracy more than the original one. The new idea for K-medoid algorithm overcomes the deficiency of existing medoid. It initially computes the initial centroids k as per the necessity of user and then provides improved and efficient cluster with no sacrifice on accuracy. It generates steady clusters in order to get better accuracy. It also minimizes the mean square error and improves the quality of clustering, reduces the number of iterations and works on reducing time complexity. The improved k-Medoid clustering algorithm will have accuracy greater than the original one.

Keywords: clustering, K-medoid, data mining, Large data sets