Effectiveness of Internal Evaluation Metrics for Community Detection Based on Clustering

Document Type

Conference Proceeding

Source of Publication

Lecture Notes in Networks and Systems

Publication Date

1-1-2024

Abstract

The exploration of complex networks and the arrangement of communities is a widely researched topic across various fields, reflecting research interest in a multitude of domains. Clustering algorithms have emerged as a prominent tool for community detection, gaining considerable attention in recent decades. To assess the effectiveness of clustering algorithms, various evaluation metrics are employed, including internal, external, and relative metrics. In this paper, the effectiveness of several partitional clustering algorithms is analyzed to identify communities. The algorithms reviewed include graph-based, centroid-based, and modal-based algorithms, which were tested on various datasets. The study’s primary aim is to determine how accurate and reliable internal evaluation metrics are for community detection through clustering. The study’s findings reveal that the k-means algorithm excelled in silhouette score and sum of squared error evaluation, while affinity propagation outperformed others in terms of the davies-bouldin index and adjusted mutual information. These results can provide valuable guidance and support in the domain of community detection, aiding researchers in achieving more accurate and effective analyses of complex network structures.

ISBN

9789819983230

ISSN

2367-3370

Publisher

Springer Nature Singapore

Volume

839

First Page

65

Last Page

75

Disciplines

Computer Sciences

Keywords

Adjusted mutual information, Clustering, Community detection, Complex networks, Davies-bouldin index, Partitional clustering, Silhouette score, Sum of square error, Un-directed graph

Scopus ID

85189503499

Indexed in Scopus

yes

Open Access

no

Link to Full Text

Share

COinS