Data-Driven Machine Learning Hazmat Incident Duration Prediction With Random Forest

Efficient management of Hazardous Materials (HAZMAT) transportation incidents is crucial for minimizing their impact on public safety and the environment. This study, leverages Random Forest, an advanced ensemble machine learning technique, to analyze and predict the duration of HAZMAT incidents. Random Forest's robust capabilities in handling non-linear relationships and complex variable interactions make it a superior predictive tool compared to traditional models. Our research utilizes a comprehensive dataset of 121 recorded HAZMAT incidents in Tennessee from 2015 to 2022, integrating variables such as material nature, crash severity, large vehicle involvement, road blockage, towing requirements, and weather conditions. By applying Random Forest, we emphasize feature importance, identifying key determinants like crash nature and vehicle size, and elucidate their impact on incident durations. The model's predictive accuracy is demonstrated through evaluation against actual incident data, yielding a Mean Absolute Error of 136, Mean Squared Error of 26,274, and an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">{2}</sup> Score of 53%. While the R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">{2}</sup> score indicates moderate predictability, it aligns with findings from prior studies, which range from 50% to 60% for similar incident duration models, providing a meaningful comparative baseline. The findings highlight Random Forest's efficacy in providing detailed insights and reliable forecasts, significantly aiding strategic emergency response planning, infrastructure development, and HAZMAT handling protocols. This study showcases how integrating machine learning with data analytics propels HAZMAT incident management toward a data-driven future, optimizing resource allocation and enhancing safety in hazardous materials transportation.