A Literature survey based on DBSCAN algorithms

Data clustering is set of techniques for analyzing data. Data mining comes in handy when tried to find patterns and attributes from a large amount of data. There have been various clustering algorithms that are used for this purpose. Clustering is an unsupervised learning technique that groups similar points together. The points in a single cluster are similar to one another and dissimilar to the points in other clusters. There are various of clustering such as partitioning clustering, density-based clustering, hierarchical clustering, etc. In this paper, researchers have tried to cover one of the density-based clustering algorithms which is Density Based Spatial Clustering of Application with Noise (DBSCAN) algorithm. The main concept of density-based clustering is to form the clusters on basis of dense and non-dense regions. DBSCAN is density-based clustering that is used for spatial clustering on noise and outlier points. It performs well on outliers and noise making it a very efficient algorithm. But it has also many disadvantages: - 1) Performs poorly when there is varied density in a dataset 2) Large data as its complexity is higher 3) Its output is dependent on user parameters which is an undesirable trait for any algorithm. RNN-DBCAN, Grid DBSCAN, MR DBSCAN, TSF DBSCAN, KNN DBSCAN, µDBSCAN algorithms are all attempts to improve the basic DBSCAN algorithm in order to address these shortcomings. In this paper, researchers look at the various DBSCAN algorithms that have been suggested so far. These modifications are analyzed rigorously, and their drawbacks are stated.