Survey of density based clustering algorithms and its variants

Abstract

Clustering technique is a unsupervised machine learning technique in the domain of data mining. Many of the clustering techniques are inherently sensitive to the input parameters. Different clustering techniques works differently for different types of the input datasets. Among all different varieties of clustering techniques, DBSCAN is one of the most important clustering technique whose working principle based on the density estimation while forming the clusters of the input dataset points which is basically used for spatial datasets of random shapes and sizes. It also eliminates the noise during the clustering formation process with a worst case run-time complexity of O(n <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">A</sup> 2). DBSCAN technique also produces a bad result for varied density datasets. In this paper we have discussed about different density based clustering techniques along with DBSCAN, its variants and some of its modified algorithms with respect to their input parameters and running time complexities. Also we have presented the comparison analysis of all the different variants of DBSCAN algorithms over different benchmark datasets for computing various measures.