An alternative parameter free clustering algorithm using data point positioning analysis (DPPA) – comparison with DBScan

S. M.F.D.Syed Mustapha, Zayed University

Document Type

Article

Source of Publication

International Journal of Innovative Computing, Information and Control

Publication Date

12-1-2023

Abstract

DBSCAN is one of the most popular clustering algorithms that could handle clusters which have characteristics of arbitrary shape, multiple densities and noises. How-ever, its accuracy depends on the right selection of the two parameters, MinPts and Eps. There have been numerous research works to overcome this issue by developing parameter free clustering algorithm. We propose a clustering algorithm which uses Data Point Positioning Analysis (DPPA) to analyze the relationship of each point to all points based on two nearest neighbor concepts, namely 1-NN and Max-NN. The algorithm is applied on 13 benchmark datasets that have been applied in many clustering algorithms with three-dimensional data and subsequently on higher dimensional data with sixteen attributes. The performance of the algorithm is visually compared with the three-dimensional graph plotting at various angles to determine the actual number of clusters. For the higher dimensional data, Silhouette coefficient is used to measure the performance. For both experimental results, the DPPA algorithm is compared against DBSCAN. The results show that the DPPA algorithm is comparable to the performance of DBSCAN algorithm such that it manages to detect arbitrary cluster shapes, identify the number of clusters and manage the data sets with noises.

ISSN

1349-4198

Volume

19

Issue

6

First Page

1805

Last Page

1825

Disciplines

Computer Sciences

Keywords

Clustering algorithm, DBSCAN, Parameter free clustering algorithm, Unsupervised learning

Scopus ID

85176778481

Indexed in Scopus

yes

Open Access

no

Link to Full Text

Share

COinS