Nested Structure of Citation Anomalies

Academic citations play an essential role in assessing scholarly impact, prompting the development of advanced metrics alongside the expansion of large academic databases. Conventional citation analyses frequently consider oversimplified citation structure, suggesting generally its uniformity, including excessive self-citations or citation cartels. This paper focuses on anomalous citation patterns appearing at multiple levels, ranging from large-scale network irregularities, such as isolated citation clusters and bursts, to finer-grained distortions, including manipulated individual citations. The study introduces a multi-level approach to citation anomaly detection. The proposed method identifies nested anomaly structures by integrating global network structures with localized citation patterns. Through hierarchical analysis, progressing from macro to micro levels, the proposed approach uncovers irregularities that single-level analyses might miss, offering a more precise and comprehensive evaluation of research influence. A case study of the CoraFull dataset demonstrates the effectiveness of the proposed model in detecting and analyzing nested anomaly structures.