Analysis of Density-Based Clustering Algorithm Variants on Real-World and Gaussian Motion Datasets

Clustering is an unsupervised machine learning technique that segregates the data points into different clusters which have similar properties. Density-based Spatial Clustering of Applications with Noise (DBSCAN) is a density-based clustering algorithm which works on the principle that a data sample belongs to a cluster if it has a required number of neighbours within a specified radius. DBSCAN produces hard clustering boundaries and performs quite poorly in case of real-world datasets. In recent years, many fuzzy variants for the algorithm have been proposed by researchers which aim to provide soft clustering boundaries thereby allowing more data points to be a part of a cluster with varying membership values. The fuzzy variations tend to perform well on these datasets. In this paper, a comparison of results of classic and fuzzy DBSCAN on four real world datasets has been done along with analysis of the results of streaming fuzzy DBSCAN variation on streaming data.