Recent studies report that few journals are adopting unethical citation practices to inflate Impact Factor (IF) artificially. “Clarivate Analytics” has started to blacklist such journals since 2006. As reported in the literature, evaluation of journals individually, to detect anomalies from vast and dynamically changing citation network is not efficient. The primary purpose of this work is to define a diverse feature set that can identify such cases of extreme outliers and reason them. The sample size is narrowed down using an unsupervised clustering algorithm in the absence of a labeled training dataset. Next, time-series IF data is analyzed to detect point outliers. Furthermore, microscopic features are identified to reason them. Results reflected from the F-value after ANOVA analysis reveals that geometrical patterns (self-loop, pairwise and group mutual-citation) among journals, an abrupt increase in the paper count of donor and corresponding IF inflation of recipient are some of the essential features. Microscopic features include social factor (calculation of revised IF after removing directed self or mutual-citation), impact of the field of study, impact of publication house and author factor that includes author self and mutual-citation. The significance of this work is to ensure that the quality of a journal is withheld without compromising research integrity by controlling or auditing individual features periodically.

中文翻译：

---

期刊中引文模式违规行为的识别与分析

最近的研究报告表明，很少有期刊采用不道德的引用做法来人为地增加影响因子（IF）。自2006年以来，“ Clarivate Analytics”已开始将此类期刊列入黑名单。如文献所报道的那样，对期刊进行单独评估以检测庞大且动态变化的引文网络中的异常情况效率不高。这项工作的主要目的是定义一个可以识别极端异常情况并对其进行推理的多样化功能集。在没有标记训练数据集的情况下，使用无监督聚类算法缩小样本量。接下来，分析时序IF数据以检测点离群值。此外，识别出微观特征以对其进行推理。结果由F反映ANOVA分析后的-值显示期刊之间的几何模式（自环，成对和小组互引），捐助方论文数量的突然增加以及接收方相应的IF膨胀是其中的一些基本特征。微观特征包括社会因素（除定向自或相互引修订后的IF计算），研究领域的影响，出版机构的影响和作家的因素，其中包括了作者的自我和相互引。这项工作的意义在于通过定期控制或审核各个功能，确保在不损害研究完整性的前提下保留期刊的质量。