The construction industry faces persistent challenges with workplace injuries, underscoring the need for advanced, data-driven safety strategies. In this study, we propose a predictive framework that uses machine learning to estimate workplace injuries and classify risk levels across construction sectors. A Neural Network model, a Random Forest model, and Linear Regression were developed using Orange Data Mining software to predict injury occurrences and classify risk levels. The injury rate was calculated using the standard OSHA formula, enabling normalization across different workforce sizes. The model’s performance was evaluated by comparing predicted and actual injury values, with error analysis used to assess accuracy. The results indicate that machine learning can effectively identify low, medium, and high-risk categories, particularly in sectors such as transportation and warehousing, which show consistently elevated injury rates. The model shows weak potential in classifying extremely high-risk cases. This advances data-driven safety management by integrating predictive analytics into construction safety practices. The findings support the use of artificial intelligence tools to enhance decision-making, reduce workplace injuries, and improve overall occupational safety performance.