Anomaly detection and community detection in networks

Abstract Anomaly detection is a relevant problem in the area of data analysis. In networked systems, where individual entities interact pairs, anomalies are observed when pattern interactions deviates from patterns considered regular. Properly defining what regular entail relies on developing expressive models for describing interactions. It crucial to address anomaly networks. Among many well-known networks, latent variable models—a class probabilistic models—offer promising tools capture intrinsic features data. this work, we propose generative approach that incorporates domain knowledge, i.e., community membership, as fundamental model behavior, and thus flags potential deviating pattern. fact, membership serves building block null identify interaction patterns. The structural information included through variables parameter. algorithm aims at inferring these parameters then output labels identifying network edges.