Bayesian Modeling Applied to Risk Estimation in the Road Cargo Transportation

Resumo

Road cargo transport (RCT) is vital to Brazil’s economy, accounting for 61.1 % of all cargo movements. Nevertheless, the sector is exposed to accidents, theft, and infrastructure damage, largely driven by road conditions, driver behavior, and cargo characteristics. While prior studies have tended to examine individual hazards in isolation, they seldom model the interdependence among these factors. This study therefore develops a probabilistic framework based on Bayesian Networks (BNs) to assess RCT risks holistically, capturing critical interactions and supporting managers in devising mitigation strategies. The research design comprised a systematic literature review, a Delphi exercise with subject‑matter specialists to prioritized 15 risk factors (excluding “Traffic Conditions”), BN construction, and validation through sensitivity analysis and scenario simulation. Twenty simulations were run - ten optimistic and ten pessimistic - followed by a survey of 105 experts to corroborate the findings. “Driver Profile,” “Cargo Type,” and “Vehicle Condition” emerged as the most influential determinants of accidents, whereas “Cargo Type” and “Driver Profile” most directly affected theft. Pessimistic scenarios raised accident probabilities to 71–72 % and theft probabilities to 80–94 %, whereas optimistic scenarios lowered these risks to 5–7 % and 13–41 %, respectively. Expert validation indicated a 70 % agreement with the model’s outputs. The proposed BN model enables dynamic assessment of RCT risk probabilities and highlights priority areas for mitigation. The study advances the modelling of complex logistics systems and underscores the efficacy of Bayesian Networks in contexts characterized by uncertainty.