<![CDATA[Traffic crashes remain a critical safety challenge, with Indonesia experiencing 73,446 fatalities annually. This study develops an integrated Z-Score and Bayesian Network framework to analyze causal interactions between human and environmental factors influencing crash severity on toll roads. Z-Score analysis of 450 crash records (2022–2025) identified five statistically significant blackspot segments, with KM 430–431 exhibiting the highest concentration (Z = 4.036, n = 91). A Bayesian Network model constructed using K2 structure learning and Expectation-Maximization parameter estimation achieved 86.2% classification accuracy, surpassing previous international applications (78–82%). Conditional probability analysis revealed that straight-downhill segments exhibited 3.3-fold higher fatal crash probability than straight-level segments (0.083 vs. 0.025), while night-time conditions increased fatal risk by 57%. Sensitivity analysis demonstrated that crash type (weighted index = 0.282) and accident cause (0.214) exerted strongest influence on severity outcomes. Human error constituted 83% of crashes but showed moderate sensitivity, indicating that severe outcomes emerge from interactions between human factors and adverse conditions rather than isolated factors. Findings support prioritizing enhanced lighting and speed management on curved-downhill segments during night-time, alongside rear-end collision prevention strategies. This validated framework enables evidence based, proactive crash management and intervention prioritization for toll road safety in developing countries.]]>
Copyrights © 2026
