Analysis of change in coordinate system on clustering
Manish Joshi, Yogita Patil · 2016
Clustering is one of the data mining techniques used in a knowledge discovery process. It is assumed that a good representation of data points may yield good clustering results [6]. This paper discusses the effect of the coordinate system on the clustering. In this paper, we propose a density based clustering approach to group objects represented using Polar coordinate system. The experiment is carried out on different datasets. To evaluate the goodness of cluster result we have used internal and external validity measures. In most of the cases, data points representation using conventional Cartesian coordinate system results in better clustering performance as compared to clustering obtained from same data represented using Polar coordinate system. However, it is observed that for certain Data Sets our proposed density based approach on Polar coordinates clustering results outperform conventional approach. Hence, we can conclude that an appropriate representation of data points may yield more appropriate clustering results.