Centrality-based procedures for monitoring directed social networks

Anomaly detection in social networks based on the degree corrected stochastic block model (DCSBM) has primarily focused on identifying structural changes in network connectivity.However, as social networks become increasingly complex, it is important not only to detect anomalies but also to characterize the direction of changes in network behavior.In this paper, we propose statistics for anomaly detection that incorporate directional changes in node propensity.To further characterize node-level behavior, the proposed approach utilizes centrality measures.Based on the proposed statistics, we develop a monitoring procedure using an exponentially weighted moving average (EWMA) chart.Through simulation studies, we evaluate the performance of the proposed procedure under various node propensity change scenarios.The results demonstrate that the proposed procedure efficiently detects directional anomalous changes in social network environments.