Establishment of Risk Management Groups in Construction Based on Workers' Age and Accident Probability Using Unsupervised Learning

The construction industry is currently experiencing increased risks due to the aging workforce. As workers age, their physical capabilities often decline, leading to an increased likelihood of accidents. Despite this known correlation, no established standards exist to assess and manage the risks associated with workers' age. This study aims to establish quantitative risk management groups based on workers' age and accident probability, providing a structured framework for age-specific safety strategies in construction. To address this gap, this study systematically assessed accident rates based on workers' age and identified risk management groups using a quantitative approach. The study began with data collection from 441 construction sites in Korea, encompassing 1.7 million workers and 2,460 accidents. Next, accident rates were calculated by worker age and categorized by construction project types, including residential, commercial, infrastructure, and plant projects. Using k-means clustering, a widely used machine learning technique for grouping data based on similarities, workers were grouped into risk management categories based on their age. Statistical validation confirmed the reliability of these clusters, demonstrating significant differences in accident rates across groups and project types. Notably, four risk management groups were identified for each project type, except for plant projects, which formed three distinct groups. These findings underscore the elevated risks faced by older workers and offer a structured, data-driven approach for safety decision-making. By providing project-specific insights, this study enables the implementation of targeted safety interventions, such as enhanced monitoring, tailored training programs, and resource allocation for high-risk groups. This framework offers decision-makers practical tools to enhance safety management and reduce accident risks effectively.