Approach for identifying unsafe road sections in the Republic of North Macedonia

The influence of road and environmental characteristics on traffic safety is a key aspect of the analysis of traffic accidents. To gain a more detailed understanding of this relationship, this study analyzes the factors that contribute the most to their frequency and severity, with an emphasis on their identification, quantitative assessment, and potential mitigation. The research focuses on the analysis of 161 sections of the main road network in the Republic of North Macedonia, with a total length of approximately 1300 km, using data related to road geometric characteristics, pavement conditions, vertical and horizontal signage, climatic influences, and traffic intensity. To assess their impact on the weighted accident index (Wi), multiple statistical and machine learning methods are applied, including correlation analysis and algorithms such as AdaBoost, Random Forest, Bagging Regressor, and Gradient Boosting, along with validation techniques. Thorough data processing and analysis enable the detection of critical factors with the greatest impact on safety, leading to the development of a methodological approach for predicting hazardous segments of the road network. The obtained results and the developed model can serve as effective tools for enhancing existing strategies for road safety assessment, allowing timely planning of interventions and reducing the risk of traffic accidents. This research represents a step towards the systematic identification of factors contributing to traffic accidents and provides a scientifically grounded approach to improving road safety measures.