Performance Comparison of Two Algorithms for Arbitrary Shapes Clustering

Discovering clusters with arbitrary shapes has attracted a large number of researchers, due to its importance in analyzing real-world datasets. DENCLUE is an efficient density-based algorithm that provides a compact mathematical definition of clusters with arbitrary shapes. Several variants of DENCLUE have been proposed to enhance its performance, including DENCLUE 2.0. This study aims to discuss the difference between the two variants, DENCLUE 1.0 and DENCLUE 2.0. This will allow for future improvements on both variants and determining the best variant for specific types of datasets. The Adjusted Rand Index measure is used to evaluate the difference between the clustering results of both algorithms. The experimental results conclude that DECNLUE 1.0 outperforms DENCLUE 2.0 in discovering clusters with arbitrary shapes. Specifically, in datasets that contain clusters with multiple modes.