A clustering approach for classifying scholars based on publication performance using bibliometric data

Abstract

This study introduces a clustering framework that effectively evaluate scholars’ publication performance by utilizing cluster analysis and bibliometric data. In order to capture the various aspects of scholars’ publication characteristics, our proposed framework integrates four distinct features, namely “APIR” which represents Academic age, Productivity, Impact, and Recency. The proposed framework is implemented in a case study focusing on Turkish academia, utilizing a dataset comprising 13,070 scholars from 24 diverse academic divisions across 30 Turkish universities. Cluster analysis yields seven groups of scholars with diverse publishing characteristic based on APIR features and these obtained clusters are profiled as “freshmen”, “stagnant impactful mids”, “rising stars”, “stagnant and non-prolific juniors”, “stagnant impactful seniors”, “super stars”, “currently active and prolific seniors”. To enhance the cluster analysis results, additional cross analysis is performed based on scholars’ certain demographics such as affiliating institutes, divisions, academic titles, and PhD qualification. Scholars in clusters with superior publication performance are often affiliated with top-ranked universities and have academic backgrounds in the fields of Medicine, Engineering, and Natural Sciences. Practically, generated scholar segments and analysis based on these scholar profiles can serve as useful input for policy makers during having decisions about recruitment, promotion, awarding and allocation of funds.

Keywords

• Data mining
• Cluster analysis
• Bibliometric indicators
• Scholar profile
• Scholar-level evaluation
• Publication performance