DefinePK

Abstract

Data mining is a technology that collects and search a bulk of data from database to discover relationship among data. It is an application that view data from different angles and group it into information that is useful in many perspectives. There are different types of clustering methods that used to grouping the generated data sets such as K-means etc. K-means algorithm is a centroid based technique and has input parameter as k. This technique has two restrictions such as k-means value selection and centroid selection i.e. the size of cluster is assigned by manually and the centroid value is selected by randomly. These two parameter impacts on the clustering performance massively. Another metric such as distance metric also have impact on choosing the cluster also presented. This paper presents powerful K-means (PKM). To show the performance of the proposed algorithm various set of dataset have been taken. That has been applied on traditional K-means and proposed algorithm. The experimental result shows proposed algorithm gives better result when compared to traditional k-means.