A Hybrid Approach for Community and Anomaly Detection in Social Networks

Detecting communities and anomalies in social networks is critical for understanding complex network structures and identifying irregular behaviors. Existing methods often address these tasks separately, overlooking their interdependence. This study addresses this gap by proposing a hybrid approach that integrates modularity-based community detection with structural and attribute-based similarity measures for anomaly detection. The framework leverages modularity optimization to identify cohesive communities and employs cosine similarity to evaluate attribute alignment for detecting anomalies. Evaluations on synthetic and real-world datasets, including the Cora and Cite seer citation networks, demonstrate the framework’s superior performance over state-of-the-art techniques, achieving higher precision, recall, and F1 scores. These findings underscore the method’s robustness, scalability, and potential for applications in social network analysis, fraud detection, and security.