Change-Point Detection in Dynamic Networks with Missing Links

Unveiling Hidden Shifts: Detecting Change Points in Dynamic Networks with Missing Links Dynamic networks are ubiquitous in a world increasingly driven by interconnected systems from social media platforms to financial transactions, transportation systems, and cybersecurity infrastructures. Unlike static entities, these networks evolve, often undergoing abrupt structural changes. Detecting these changes is essential as they can signal critical events, such as cyberattacks or shifts in user behavior. The challenge becomes even greater when dealing with incomplete data, a common reality in large-scale networks. Missing links, from nonresponses or privacy concerns, create uncertainty, complicating the identification of structural changes. The paper “Change-Point Detection in Dynamic Networks with Missing Links” tackles this issue with a robust framework based on the matrix CUSUM test statistic. This novel methodology is adaptable to expanding networks and resilient to missing data. It detects structural shifts with minimax optimality, ensuring theoretical robustness. Extensive simulations and real-world examples reveal that the proposed approach outperforms existing methods. It provides a valuable tool for monitoring dynamic networks, detecting fraud, and spotting emerging trends.