Road traffic accidents remain a major problem within Indonesia's road transport system, resulting in significant social and economic impacts. One of the ways to enhance road safety is undertaken through infrastructure risk assessments using a star rating approach. Currently, these assessments are conducted via two methods: the International Road Assessment Programme (iRAP) Star Rating method, which is an international standard, and the Road Functional Feasibility Assessment (ULFJ) Star Rating method, developed specifically for Indonesian conditions. This study aims to compare the road safety assessment results of these two methods, as well as to analyse their accuracy through ground truth validation and the sensitivity of infrastructure attributes to changes in risk values. The novelty of this research lies in the ground truth validation and attribute sensitivity testing of each method in assessing road safety risks. The analysis was conducted on the Pekutatan–Antosari national road section in Bali Province, using road segments as the units of analysis. Comparisons were made regarding the Star Rating Score (SRS) and star categories, with accuracy tested against actual accident data representing collective and personal risk. Furthermore, sensitivity testing was performed through attribute change scenarios to identify the factors most influential to road safety risk levels. The results indicate differences in risk values and star categories between the iRAP and ULFJ methods, influenced by variations in attribute structure, Crash Modification Factors (CMF), and approaches to traffic characteristics and road users. While the iRAP method excels in analytical depth and sensitivity to physical road attributes, the ULFJ method demonstrates a higher level of alignment with actual accident data and is more responsive to operational attributes. These findings suggest that the ULFJ method is more representative within the context of national technical standards for assessing road safety risks on the Pekutatan–Antosari section.
Copyrights © 2026