Developing a novel algorithm for comparing cluster patterns in networks on journal articles during and after...

Area of Science:

• Bibliometrics and network analysis
• Information science
• Data mining

Background:

• Bibliometrics relies on cluster analysis for large datasets.
• No prior research used cluster-pattern algorithms for comparing two clusters.
• This study addresses this gap by developing and applying a novel algorithm.

Purpose of the Study:

• To create a cluster-pattern comparison algorithm (CPCA) for bibliometric analysis.
• To apply CPCA to analyze clusters of countries, institutes, departments, authors (CIDA), and keywords.
• To compare these clusters during and after the COVID-19 pandemic.

Main Methods:

• Analyzed 9499 (2020-2021) and 5943 (2022-2023) articles from the Journal of Medicine (Baltimore).
• Developed a cluster-pattern-comparison algorithm (CPCA) using similarity coefficients, collaborative maps, and thematic maps.
• Compared follower-leading clustering algorithm (FLCA) with eight other algorithms for validation.

Main Results:

• Similarity coefficients for CIDA entities and keywords were 0.73, 0.35, 0.80, 0.02, and 0.83.
• Identical patterns (>0.70) were found in country, department, and keyword clusters.
• Dissimilar (<0.5) and different (<0.3) patterns were observed in institute and author clusters.

Conclusions:

• The study successfully developed and applied CPCA for bibliometric cluster pattern analysis.
• Identical patterns exist in country, department, and keyword clusters, but not in institute and author clusters.
• CPCA provides a framework for future bibliometric studies to compare diverse cluster patterns.