A Novel Multistage Approach to Community Detection in Social Networks Based on Fuzzy DBSCAN Clustering and Reptile Search Optimization

Abstract

Community detection is a fundamental problem in the analysis of complex and social networks. It aims to identify groups of nodes whose internal connections are denser than their connections with other parts of the network. The primary challenges in this field are the unknown number of communities and the large number of nodes in real-world networks. This study proposes a novel multistage approach that combines the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm with the Reptile Search Algorithm (RSA). In the first stage, the network is initially clustered using DBSCAN and features extracted from the nodes, particularly their neighbor lists. In the second stage, RSA identifies the final communities through three phases: initialization, exploration, and exploitation. The objective function is defined based on two criteria: modularity (Q) and conductance (Cond). The proposed method was evaluated in MATLAB using eight real-world datasets: ca-GrQc, ca-HepPh, ca-HepTh, ca-AstroPh, ca-CondMat, Zachary’s Karate Club network, the American college football network, and the dolphin social network. The experimental results demonstrate that the proposed method outperforms existing methods in most cases, achieving higher values for normalized mutual information (NMI) and modularity (Q).