KMK based hybrid approach for the performance estimation in case of diabetes data

Abstract

In this paper k-means clustering algorithm has been used with k-points (KMK) selection. It has been applied on the PIMA Indian diabetes dataset. It has been used for distance estimation, centroid selection, effect of data size variations and for the analysis of the complete record. The cluster section has been found to be improved based on k-point selection. It has been used for the assignment of initial centroid. The results indicate that the KMK algorithm is capable in the improvement of centroid selection and distance measures in the assignments of data points. It is due to the better centroid selection mechanism by k-points selection based on the weight measures from the selected dataset. So, the obtained clusters are better in comparison to k-means.