OPTIMAL LINK PREDICTION IN COMPLEX HUMAN NETWORKS VIA STRUCTURAL PATTERN ANALYSIS

Abstract

This study investigates optimal link prediction methods in complex human networks by applying eight similarity-based algorithms across five real-world human interaction datasets. Networks were converted into adjacency matrices, with links removed for testing. Each method was evaluated using AUC scores to assess predictive performance. Results showed that the Resource Allocation Index (RAI) consistently outperformed other techniques, particularly on large and complex networks, highlighting its effectiveness for predicting missing or future connections in human-centric systems.