Mapping neural network research in education through bibliometric analysis

Neural networks are reshaping educational research and practice; yet evidence syntheses often treat them as part of a broader “AI in education” landscape, obscuring neural-network-specific trajectories, contributors, and themes. This study addresses that gap by providing a focused, methodologically transparent bibliometric mapping of neural network research in education. We analysed 704 Scopus-indexed journal articles published between 2010 and 2023, following PRISMA 2020 for identification, screening, and inclusion. Bibliometric performance analysis and science mapping were conducted using VOSviewer (for co-authorship, co-occurrence, and density visualisations) and the R-based Bibliometrix package (for descriptive indicators and thematic evolution). Network construction applied fractional counting and association-strength normalisation, with robustness checks using alternative thresholds. Findings show accelerated publication growth from 2019 onward, culminating in the highest output in 2023. The United States, the United Kingdom, and China lead productivity, while key institutions and author clusters function as collaboration hubs. Keyword co-occurrence reveals five dominant thematic clusters: (1) AI and machine learning foundations, (2) deep learning and neural network architectures, (3) learning analytics and personalised learning, (4) ethics, fairness, and explainability, and (5) higher education digital transformation. Thematic evolution indicates a shift from early automation-oriented work toward applied learning analytics and, more recently, governance concerns such as academic integrity and responsible AI, alongside emergent terms linked to generative AI. These results provide a replicable baseline for tracking intellectual structure, collaboration patterns, and ethical priorities in the field of neural network scholarship in education. We recommend strengthening interdisciplinary teams, expanding Global South participation through open infrastructure and partnerships, and prioritising transparent, fairness-aware designs when translating neural network research into educational policy and practice.