An explainable artificial-intelligence-based approach to investigating factors that influence the citation of papers

Keywords Bibliometrics; Citations; Machine learning; Big data; SHAP Highlights * This study examines 14 factors that can influence the citation of papers. * CatBoost model and SCOPUS dataset are used to examine the influences. * SHAP interprets the model and suggests how the factors contribute to the citation. * The results show that selecting the right journal/conference is the most important. Abstract The number of citations is often used to estimate the impact of a study. Previous studies have investigated what factors of publications affect citations and how they affect citations. However, the findings of the studies were unable to reach a consensus because of the limited sample size, domain, and measurement. This study reviewed previous studies that addressed factors influencing citations and then identified 14 measurable factors. Approximately 33 million publications from the Scopus database were used to train and validate a CatBoost model. A SHAP framework was used to interpret the trained model by focusing on how salient factors affect the number of citations. The results showed that the year is a significant factor affecting the citation but not the priority factor. A publication source was presented as the most important factor contributing to the citation. Several implications and strategic approaches to maximizing the impact of a study were discussed. Author Affiliation: Future Technology Analysis Center, Korea Institute of Science and Technology Information, 66 Hoegiro, Dongdaemun-gu, Seoul 02456, South Korea Article History: Received 17 February 2022; Revised 30 June 2022; Accepted 18 August 2022 Byline: Taehyun Ha [taehyunha@kisti.re.kr]