Analyzing author collaborations by developing a follower-leader clustering algorithm and identifying top co-authoring countries: Cluster analysis

Abstract

Background: This study aimed to explore suitable clustering algorithms for author collaborations (ACs) in bibliometrics and investigate which countries frequently coauthored with others in recent years. To achieve this, the study developed a method called the Follower-Leading Clustering Algorithm (FLCA) and used it to analyze ACs and cowords in the Journal of Medicine (Baltimore) from 2020 to 2022. Methods: This study extracted article metadata from the Web of Science and used the statistical software R to implement FLCA, enabling efficient and reproducible analysis of ACs and cowords in bibliometrics. To determine the countries that easily coauthored with other countries, the study observed the top 20 countries each year and visualized the results using network charts, heatmaps with dendrograms, and Venn diagrams. The study also used chord diagrams to demonstrate the use of FLCA on ACs and cowords in Medicine (Baltimore). Results: The study observed 12,793 articles, including 5081, 4418, and 3294 in 2020, 2021, and 2022, respectively. The results showed that the FLCA algorithm can accurately identify clusters in bibliometrics, and the USA, China, South Korea, Japan, and Spain were the top 5 countries that commonly coauthored with others during 2020 and 2022. Furthermore, the study identified China, Sichuan University, and diagnosis as the leading entities in countries, institutes, and keywords based on ACs and cowords, respectively. The study highlights the advantages of using cluster analysis and visual displays to analyze ACs in Medicine (Baltimore) and their potential application to coword analysis. Conclusion: The proposed FLCA algorithm provides researchers with a comprehensive means to explore and understand the intricate connections between authors or keywords. Therefore, the study recommends the use of FLCA and visualizations with R for future research on ACs with cluster analysis.

Trends and hotspots related to traditional and modern approaches on acupuncture for stroke: A bibliometric and visualization analysis

“…The FLCA algorithm [54–67] is straightforward but needs refinement. Improvements could include giving precedence to individuals with higher WCD for maximum connection and incorporating a stop criterion for relationships within the AC network.…”

A modern approach with follower-leading clustering algorithm for visualizing author collaborations and article themes in skin cancer research: A bibliometric analysis

“…Traditional bibliometrics, heavily relying on software like CiteSpace [35] and VOSview, [36] present cluttered results with numerous edges and nodes in networks (e.g., 26 Figures and 15 Tables in a bibliographical study [37] ), limiting structured layouts and concise cluster analysis. These drawbacks highlight the need for improvements that combine visuals, with the DDPP model, [11,12] and the FLCA algorithm [13,14] to enhance the quality and depth of SC research. Through which, overcoming these limitations would greatly contribute to a more comprehensive understanding of SC within the field of bibliometrics.…”

“…[7,8] Despite the availability of over 125,041 articles on SC indexed in PubMed, [9] traditional bibliometric methods have only been applied to a mere 6 documents. [10] The utilization of modern approaches such as the DDPP model (descriptive, diagnostic, predictive, and prescriptive analytics) [11,12] and the FLCA (follower-leading clustering algorithm) [13,14] in bibliographical studies has been limited in this context. It is challenging to explore author collaborations and article themes in SC research.…”

Thematic maps with scatter and 4-quadrant plots in R to identity dominant entities on schizophrenia in psychiatry since 2017: Bibliometric analysis

“…The follower-leading cluster algorithm (FLCA) [44][45][46] was applied to classify clusters. A collaborative map with 4 quadrants (namely, emerging, niche, co-work (or co-word), and motor by nodes' edge degree and frequency degree, respectively, in network) was made in R programing language [47] generated on a R platform, [48] based on country-based author collaborations and keyword cooccurrences, respectively.…”

“…One hundred top-cited articles related to schizophrenia in psychiatry (T100SCHIZ) [49] were extracted from the current study. Keywords Plus in WoS was used to cluster research domains using the FLCA, [44][45][46] as displayed in social network analysis (SNA). [50][51][52] Keywords with the highest weighted score in the respective cluster using SNA in the previous section would be assigned as the representative of the cluster.…”