Ant Colony Optimization for Link Prediction in Online Social Networks: A Swarm Intelligence Approach

Link prediction in online social networks (OSNs) is critical with manifold applications, spanning recommendation systems, friend suggestion algorithms, and community detection. Various techniques exist, including topology-based, machine learning-based, and probabilistic approaches, for computing similarity measures between node pairs. However, prediction outcomes often fall short in accuracy and computational cost, particularly for machine learning and probabilistic methods. While topology-based prediction is favored for its lower computational cost, its accuracy remains suboptimal. This paper proposes a novel approach that integrates local topology with a swarm intelligence technique, ant colony optimization (ACO), drawing inspiration from ants’ foraging behavior. The proposed approach computes similarity metrics between prospective pairs of nodes, aiming to boost prediction based precision while reducing computational expenses. Numerous comprehensive tests are conducted on five real-world OSN datasets of varying sizes and sourced from diverse fields to assess the efficiency of the ACO-based link prediction technique. The performance of this proposed algorithm has been compared against two local topologies, two global topologies and two quasi-local based state-of-the-art topology algorithms, and its findings indicate that integration of ACO offers a promising alternative for the link prediction in OSNs, where traditional topology-based methods face challenges related to scalability and adaptability. This study lays the foundation for future research on improving link prediction algorithms in dynamic online contexts and significantly advances swarm intelligence techniques in social network analysis.