Can we predict ESI highly cited publications?

Abstract

The highly cited papers defined by Clarivate Analytics’ Essential Science Indicators (ESI) have been widely used to measure the scientific performance of scientists, research institutions, universities and countries. However, researchers have seldom studied which factors can affect a paper to be an ESI highly cited paper. The prediction of ESI highly cited papers is much less studied, too. According to the existing researches about factors influencing paper’s citations, four classical papers’ factors are chosen in this study, which are scientific impact of the first author, scientific impact of the potential leader, scientific impact of the team and the relevance of authors’ existing papers. Similar to the definition of ESI highly cited papers, we develop a new measure of papers’ scientific impact. Firstly, we get statistics properties of four factors with APS data and Nobel data in order to study four factors’ performance of ESI highly cited papers. Then, Spearman correlation and Logistic regression are applied to explore the relationship between four factors and papers’ scientific impact. At last, we try to predict highly cited papers by NN algorithms incorporating four factors. The results show that the potential leader factor plays a more important role in the short term than in the long term, while the team factor is on the contrary, more important in the long term. Interestingly, the first author factor doesn’t have an obvious effect on papers’ scientific impact among top 1%. The prediction results are better than random.