Personalised Construction Site Safety Management Using Machine Learning: A Bibliometric Review

Purpose: Construction sites have high accident rates, emphasising the need for better safety management. This study examines the current research on personalised machine learning (ML) safety management in construction through a bibliometric review. Design/Methodology/Approach: A critical realist approach was used within the historical context. Scopus served as the source for bibliographic data extraction, focusing on publications from 2014 to 2024 and using specific search phrases. Data analysis involved co-citation, citation, and keyword analyses with VOSviewer software. Research Limitation: This study examines a specific timeframe and may overlook recent developments. Relying solely on published literature may not capture all current research efforts. Findings: The review identified 90 publications, showing increasing interest in personalised ML-based safety management. The United States, China, and South Korea lead the research efforts, with notable contributions from Turkey and India. The findings highlight the potential of ML algorithms such as random forest, text mining, and deep learning for personalised safety interventions. Practical implications: This study provides valuable insights for researchers and practitioners on the current research landscape and promising opportunities to utilise ML in construction safety management. Construction professionals can leverage federated learning and edge computing for distributed and localised data processing in personalised safety systems. Social implications: This study provides insights into the development of personalised safety solutions to improve worker safety and well-being in the construction industry. Originality/Value: This study enhances the understanding of personalised ML in construction safety by showcasing potential algorithms and their ethical implications. It provides practical guidance for researchers and practitioners to support further development.