A varied density-based clustering algorithm

Discovering clusters of different sizes, shapes, and densities is a challenging duty. DBSCAN can find clusters of different shapes and sizes. But it has trouble finding clusters of different densities because it depends on a global value for its parameter Eps. Several methods have been proposed to t...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of computational science Jg. 66; S. 101925
1. Verfasser: Fahim, Ahmed
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier B.V 01.01.2023
Schlagworte:
Cluster analysis
Clustering algorithms
k-nearest neighbors
Varied density clusters
VDCA
Cluster analysis
Varied density clusters
VDCA
Clustering algorithms
k-nearest neighbors
ISSN:1877-7503, 1877-7511
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Discovering clusters of different sizes, shapes, and densities is a challenging duty. DBSCAN can find clusters of different shapes and sizes. But it has trouble finding clusters of different densities because it depends on a global value for its parameter Eps. Several methods have been proposed to tackle this problem, each method has its drawbacks. This paper introduces a new stand-alone method to discover clusters of different densities. The proposed method depends on the k-nearest neighbors to compute the local density of each object as the sum of distances to its k1-nearest neighbors, where 0 < k1 < k, it starts from any object. This object is called a cluster initiator. Any object that is reachable from a cluster initiator and has a local density similar to the local density of the cluster initiator is assigned the same cluster. So, the method requires a threshold for similarity, which will be called SR (Similarity Ratio). The proposed method discovers clusters of different densities, shapes, and sizes. The experimental results show the superior ability of the proposed method to detect clusters of different densities even with no discernible separations between them. •Discovering clusters of varied densities.•A density-based clustering algorithm based on k-nearest neighbors and local density of objects.•Handling varied density clusters with noise.
ISSN:1877-7503
1877-7511
DOI:10.1016/j.jocs.2022.101925