Evaluating the Effectiveness of Star Ratings for Road Accident Severity in Thailand's Rural Roads

Abstract: Inaccurate risk evaluation significantly contributes to the severity of accidents and injuries in Thailand. Although the International Road evaluation Program (iRAP) star rating system is used for road safety evaluation, there are still inconsistency as roads with high ratings still experience significant accidents. This indicates that the current indicators may not provide a comprehensive representation of road safety conditions in rural areas of Thailand. To address this, the study proposes using machine learning techniques to optimize the Star Rating Score (SRS) calculation based on available accident data. The objective of the research is to examine the influence of star ratings on the severity of accidents, assess the ability of crash type scores to forecast accidents, and develop a modified formula for Star Rating Scores (SRS). The study involves collecting and analyzing data from the rural road accident database, applying machine learning to improve the accuracy of star ratings, and comparing the new results with existing data. The results, obtained by correlation measures and logistic regression, indicate that there is no statistically significant association between accident severity levels and SRS in both the current and adjusted models. This suggests that SRS may not be a reliable predictor of accident severity. Moreover, the standardized indicators included in the SRS may not sufficiently capture the distinct attributes of rural roads in Thailand, emphasizing the necessity for additional improvement. Recommendations for enhancing the SRS model for rural roads are proposed, including policy implications to improve road safety in these areas.