Measuring social media activity of scientific literature: an exhaustive comparison of scopus and novel altmetrics big data

• Published in: Scientometrics Vol. 113; no. 2; pp. 1037 - 1057
• Main Authors: Hassan, Saeed-Ul, Imran, Mubashir, Gillani, Uzair, Aljohani, Naif Radi, Bowman, Timothy D., Didegah, Fereshteh
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.11.2017; Springer Nature B.V
• Subjects: Astronomy; Bibliometrics; Citation analysis; Computer Science; Data management; Data processing; Digital media; Documents; Information Storage and Retrieval; Informetrics; Library Science; Nursing; Physics; Social networks; Comparative analysis; Research evaluation; Scopus; Altmetrics
• ISSN: 0138-9130; 1588-2861
• DOI: 10.1007/s11192-017-2512-x

This paper measures social media activities of 15 broad scientific disciplines indexed in Scopus database using Altmetric.com data. First, the presence of Altmetric.com data in Scopus database is investigated, overall and across disciplines. Second, a zero-truncated negative binomial model is used to determine the association of various factors with increasing or decreasing citations. Lastly, the effectiveness of altmetric indices to identify publications with high citation impact is comprehensively evaluated by deploying area under the curve (AUC)—an application of receiver operating characteristic. Results indicate a rapid increase in the presence of Altmetric.com data in Scopus database from 10.19% in 2011 to 20.46% in 2015. It was found that Blog count was the most important factor in the field of Health Professions and Nursing as it increased the number of citations by 38.6%, followed by Twitter count increasing the number of citations by 8% in the field of Physics and Astronomy. The results of receiver operating characteristic show that altmetric indices can be a good indicator to discriminate highly cited publications, with an encouragingly AUC = 0.725 between highly cited publications and total altmetric count. Overall, findings suggest that altmetrics can be used to distinguish highly cited publications. The implications of this research are significant in many different directions. Firstly, they set the basis for a further investigation of altmetrics efficiency to predict publications impact and most significantly promote new insights for the measurement of research outcome dissemination over social media.