Comparative Analysis Review of Pioneering DBSCAN and Successive Density-Based Clustering Algorithms

Abstract

The density-based spatial clustering of applications with noise (DBSCAN) is regarded as a pioneering algorithm of the density-based clustering technique. It provides the ability to handle outlier objects, detect clusters of different shapes, and disregard the need for prior knowledge about existing clusters in a dataset. These features along with its simplistic approach helped it become widely applicable in many areas of science. However, for all its accolades, the DBSCAN still has limitations in terms of performance, its ability to detect clusters of varying densities, and its dependence on user input parameters. Multiple DBSCAN-inspired algorithms have been subsequently proposed to alleviate these and more problems of the algorithm. In this paper, the implementation, features, strengths, and drawbacks of the DBSCAN are thoroughly examined. The successive algorithms proposed to provide improvement on the original DBSCAN are classified based on their motivations and are discussed. Experimental tests were conducted to understand and compare the changes presented by a C++ implementation of these algorithms along with the original DBSCAN algorithm. Finally, the analytical evaluation is presented based on the results found.