Objectives: Road traffic crashes remain a major public health concern in Saudi Arabia, which recorded a high fatality rate of 18.5 per 100,000 population in 2021. The objective of this study is to develop machine learning models which can identify crash severity determinants and predict the severity of crashes with high accuracy.Method: This study analyzes 4,979 crashes from 2017–2022 in the Eastern Province to identify factors influencing crash severity. A multinomial logistic regression model was developed using crash severity as the dependent variable, achieving an overall prediction accuracy of 88.8%. A Multilayer Perceptron Neural Network was also developed with achieved 88.6–89.7% accuracy.Results: Significant predictors of fatal crashes included the number of people involved, peak-hour periods, autumn season, clear weather, sandstorms, and driver speeding. For injury crashes, key predictors included road geometry, surface condition, collision type, and driver distraction.Conclusion: Findings provide locally grounded insights to guide targeted road safety interventions which can help in improving the road safety situation to support the national Vision 2030’s goal of reducing fatalities below 10 per 100,000.
Minhas et al. (Thu,) studied this question.