Citation burst prediction in a bibliometric network

• Published in: Scientometrics 2022-05, Vol.127 (5), p.2773-2790
• Main Authors: Amjad, Tehmina, Shahid, Nafeesa, Daud, Ali, Khatoon, Asma
• Format: Article
• Language: English
• Subjects: Accuracy; Bibliometrics; Citation analysis; Computer Science; Conferences; Correlation analysis; Information Storage and Retrieval; Library Science; Regression analysis; Regression models; Visibility
• ISSN: 0138-9130; 1588-2861
• DOI: 10.1007/s11192-022-04344-3

In the field of computer science, both journal and conference publications are considered valuable. The popularity of an author is mostly determined by the paper’s high citations in a short time. Features that can help to attract higher visibility are not yet thoroughly investigated in the literature. This study aims to investigate the impact of the several features on received citations, for articles published in both journals or conferences. The correlation analysis and multiple linear regression models are applied to explore the strength of all related features. The study helps in finding the impact of the individual features on the number of citations both for journals and conferences, and to predict future citations. AMiner citation dataset has been used for experimental analysis. The findings of the study show that in the case of journal publications, “author first-year citations” and “author total citation” are the most important features. While, in the case of conference publications, “author total citation” is more effective as compared to other features. In the case of journal publications, the multiple linear regression model shows the coefficient of determination ( R 2 ) is 0.975 and accuracy 0.846. For the conference publications, the R 2 value and accuracy are 0.877 and 0.846, respectively.