DefinePK

Abstract

In this paper, we mainly focus on tweaking the performance of clustering by K-means for the given acquisition data. The data include a lot of attributes having different categories. Mainly the attributes are categorized into Numerical attributes and Categorical attributes. By using these attributes, data can be classified into a) Numerical data having only numerical attributes b) Categorical data having only categorical attributes c) Mixed data having both Categorical and numerical attributes. Initially, the Correlation Analysis is used for knowing the relationship among the attributes in the given data. It is exceptionally hard to discover Correlation Analysis for a tremendous measure of information. It may be conceivable of missing the traits with the tremendous measure of information. So in this paper, the correlation map is constructed for visualizing the correlated attributes by leaving irrelevant attributes among the given acquisition data. This correlated data available from the correlation map are used for tweaking the performance of K-means clustering results. For extracting the correlated data and tweaking the k-means clustering results, the Correlation based K-means Clustering (CBK) algorithm is proposed. In this paper, we mainly visualize the Clustering Accuracy and Normalized Mutual Information (NMI) among the attributes using K-means and future Correlation based k-means (CBK).