Using Predictive Analytics to Model Operators' Health and Safety Trajectory

In this cross-campus collaboration with support from industry, Professors Leonard D. Brown and Hong Cui led an investigative team to move the mining industry towards using leading indicators (pre-incident predictors) rather than lagging indicators (realized accidents, injuries, or other negative outcomes) to improve the health and safety trajectory of mine operators.

This project employed Machine Learning to better utilize safety reports to infer potential incidents and improve risk management at mine sites. Data was obtained from the Mine Safety & Health Administration’s Accidents & Injuries public dataset and segmented into common classes of injury in mining and other industries. Models were validated and tested against hundreds of operator-reported Safety Interaction Reports (SIRs). Ongoing development will increase the robustness of algorithms and datasets, enhance the web-based dashboard, and allow for scaling and deployment into a production environment. A data analytics course and workshop are planned as professional development offerings, helping to promote a pre-emptive versus reactive approach to mine worker safety.