From prediction to explanation: A machine learning and causal mediation framework for roadway crash risk with connected vehicle data

Contribution

Pairs connected-vehicle crash-risk prediction with causal mediation to investigate the pathways linking roadway conditions, driving behavior, and risk.

This paper represents the connected-vehicle and causal-explainability line of work: predictive safety models are paired with causal mediation analysis so the model output can be translated into interpretable crash-risk mechanisms.