Efficient Link Prediction in Social Networks Using a BBOA-Inspired Heuristic and Similarity-Based Machine Learning

Abstract

Social networks (SNs) play a vital role in enabling information exchange, coordination, and user connectivity across various domains. Accurate link prediction in these networks is essential for uncovering hidden relationships, identifying influential users, and enhancing the efficiency of information flow. In this study, we introduce a novel hybrid framework that combines structural similarity-based metrics with supervised machine learning algorithms to improve link prediction performance. Specifically, we extract twelve structural similarity indices and employ a Binary Butterfly Optimization (BBO) inspired heuristic algorithm to select the most discriminative features. These features are then used to train and evaluate using multiple machine and deep learning classifiers.The experimental results on two real-world social network datasets—Facebook and Enron Email-Eu-Core—demonstrate that the proposed approach significantly enhances prediction accuracy. Notably, the CNN model achieves superior results, with 98% accuracy on the Facebook dataset and 85% on the Email-Eu-Core dataset. Furthermore, to ensure model transparency and explainability, SHapley Additive exPlanations (SHAP) is applied to interpret the contribution of each selected feature. Overall, the findings highlight the effectiveness of the proposed hybrid methodology in advancing link prediction in complex social networks.