LEVERAGING STRUCTURAL PATTERNS TO IDENTIFY OPTIMAL LINKS IN COMPLEX HUMAN NETWORKS

Abstract

Link prediction in complex human networks aims to identify missing, future, or potential connections. This study evaluates eight similarity-based algorithms across five human interaction networks to determine the most effective method. Datasets were represented as adjacency matrices and split into training and testing sets, with removed links used to assess predictive performance. Similarity scores were computed during training, and predictive accuracy was measured using the Area Under the Curve (AUC) metric. Comparative results indicate that the Resource Allocation Index (RAI) consistently outperforms other algorithms, particularly in large and complex networks, demonstrating its effectiveness for link prediction in human interaction systems.