Dynamic neighborhood selection (DNS) clustering using MapReduce

Data centers has rich source of data required for analyzing and studying any type of environment. In recent days, an enormous amount of data is being collected worldwide, but difficult to process with regular statistical tools. As of now, clustering with a vast amount of data is a challenging problem. Clustering the features of a single domain from a datacenter has given birth to the need for more sophisticated statistical tools because of the sensitivity of the data. On traditional density based clustering methodologies, temporal and spatial efficiency is less. In this paper a dynamic neighborhood selection (DNS) clustering algorithm is proposed which process a vast amount of data with varying densities using the MapReduce framework which is used to parallelize the clustering technique. Based on this technique, features of individual attribute from heterogeneous domain have been clustered using correlation measures of data points and it improves temporal efficiency. The experimental results show that DNS Cluster works efficiently with varying densities and scales up well with Map Reduce framework with large data sets.