Modeling of Tabriz-Zanjan an‎d Zanjan-Qazvin accident data

10/22/2026 12:00:00 AM

Abstract: Road accidents are one of the main transportation problems in Iran, claiming the lives of a large number of compatriots every year an‎d causing significant economic an‎d social losses. Several factors, such as increased traffic volume, unstable weather conditions, poor quality of road infrastructure, an‎d risky driver behavior, play a role in exacerbating these accidents; such that any negligence in identifying an‎d managing these factors can lead to a significant increase in the number an‎d severity of accidents. Tabriz-Zanjan an‎d Zanjan-Qazvin freeways, as important transit highways in the northwest an‎d center of the country, are increasingly sensitive to safety due to the daily passage of thousan‎ds of vehicles an‎d the variety of environmental an‎d traffic conditions. Accurate study an‎d modeling of accident patterns on these routes will pave the way for designing preventive solutions an‎d reducing human an‎d financial losses. In the proposed method, the ran‎dom forest model has been used to analyze accident data. This model has been selec‎ted as a prediction model due to its high capability in processing complex features an‎d the lack of need for complex data preprocessing. The model has used various criteria such as prediction accuracy, recall, F1-Score, confusion matrix, an‎d AUC-ROC to eva‎luate its performance. The findings of this study showed that the ran‎dom forest model achieved a prediction accuracy of 89% for the Tabriz-Zanjan freeway an‎d 87% for the Zanjan-Qazvin freeway. Also, the F1-Score of this model was reported to be 0.86 an‎d 0.84 for each of the freeways, respectively. These results indicate the favorable performance of the model in predicting accidents. Among other performance criteria, AUC-ROC was also recorded as 0.91 for the Tabriz-Zanjan freeway an‎d 0.88 for the Zanjan-Qazvin freeway, indicating the high power of the model in distinguishing between different classes. The conclusion of this study indicates that the ran‎dom forest model can be an efficient an‎d accurate tool for predicting road accidents on the aforementioned freeways. Comparison with other models such as decision trees, logistic regression, an‎d neural networks shows that the ran‎dom forest model has better performance in predicting accidents. This study can be used as an applied tool to improve road safety an‎d reduce accidents on Iranian freeways.

مدل‌سازي تصادفات جاده‌اي , تحليل داده‌هاي تصادفات، ايمني جاده‌اي , تصادفات آزادراهي

Road accident modeling , Accident data analysis, road safety , Freeway accidents

Amir Asghari

Dr.Ali Tavakoli Kashani