AI-Driven Safety Analytics for Cost Reduction and Operational Efficiency in High-Risk Environments

Reducing workplace incidents is both a safety priority and a business objective because incidents create direct costs, downtime, and productivity losses. This paper presents a business-oriented safety analytics framework that predicts whether a training intervention will be effective for a given worker and then uses that prediction to prioritize training under a fixed budget. Using a structured dataset of 4,000 training records, Gradient Boosting achieved ROC-AUC of 0.916 and F1-score of 0.884 on a held-out test set. We then simulate a budget-aware policy that selects workers with the highest expected benefit, combining incident-risk proxies and predicted training success. Under scenario assumptions, the proposed policy avoids approximately $414,122 of expected incident cost at a $150,000 budget while covering 170 workers. The results show how even straightforward predictive models can support practical training investment decisions when paired with transparent scenario analysis.