Year: 2026
Type: conference-paper
Abstract: Heat-related illnesses (HRIs) pose a persistent threat to construction workers' safety.Effective risk assessment enables early detection of hazardous conditions and timely prevention of illness, yet incident logs and narrative reports are seldom leveraged for computational risk analysis.This research develops a data-driven framework to (i) identify HRIs from routine safety records and (ii) predict heat-related risk based on historical incident and environmental conditions to support planning and site controls.The approach combines text mining and supervised classification to detect heat-related cases from narratives, supported by structured incident details, including task, project context, and HRI indicators.Incident time and location are combined with environmental heat-stress indices to analyze environment-task-risk relationships.A predictive model is developed to estimate near-term risk levels for various construction activities.Results reveal clear seasonal and regional patterns, with higher risks in outdoor and heavy-civil operations during prolonged heat waves.The classification model performed well, with an F1 score of about 0.72-0.81across trade groups, and probabilistic heatrelated risk prediction offers actionable lead time for planning adjustments.The risk scores provide a foundation for potential integration into construction scheduling systems to support work-rest planning, hydration strategies, shade provisioning, and task resequencing during high-heat periods.This work provides a scalable computational framework that transforms routine safety records into predictive insights, enabling proactive, climate-aware risk mitigation in construction.
Institution: The University of Texas at San Antonio
Language: English
Subfield: Safety Research
Field: Social Sciences
Domain: Social Sciences
Open Access status: bronze
OpenAlex ID: W7165922676