Relationship between number of downloads and three journal-based metrics of 11 subject categories among 1575 Springer Nature journals

The number of downloads (NOD) is a measure of the number of accesses to (or downloads of) published articles and a subset of altmetrics. In this study, we assessed the correlation between the journal impact factor (JIF) and NOD for 11 subject categories on Springer Nature’s Springerlink to determine if there were differences in NOD among Google Scholar, Scopus (CiteScore) and Clarivate’s JIF across these subject categories, and attempted to predict NOD using JIF. From a total of 1575 journals, 1155 (73.3%) were grouped under JIF, 275 (17.5%) under CiteScore, and 145 (9.2%) under Google Scholar. Among the 1155 JIF journals, 1007 (87.2%) were subscription or hybrid journals while 148 (12.8%) were open access journals. Except for “environment”, there was a significant positive correlation between NOD and JIF for all remaining subject categories. Correlations changed slightly even after open access was removed from all categories. The Kruskal Wallis test showed significant differences in median NOD for journals with a CiteScore, Google Scholar and JIF, and this was fortified by a posthoc test (Conover p-values without adjustment). After aggregating the data of all subject categories into two sub-categories (NOD and JIF) of the 1155 journals with a JIF, finally, Adaptive Boosting performed best among eight machine learning models to predict NOD using JIF (RMSE = 84139.1; R2 = 0.9669). This research extends researchers’ understanding of the relationship between altmetrics and citations with journal metrics that are typically obtained using citations. Knowledge of a JIF can predict NOD with some permissible error.