Random Forest Based Prediction of Casualties in Expressway Accidents

Le résumé fourni par la source

Expressways, as the core arteries of modern transportation networks, have significantly enhanced travel efficiency and logistics speed, vigorously promoting regional economic integration. However, their enclosed nature, high speeds, and complex traffic environments also markedly increase the risk of traffic accidents. Therefore, effectively predicting expressway casualties and preventing traffic accidents is of paramount importance. To explore the application prospects of machine learning algorithms in this field, this paper first conducts data analysis on the 2019 UK expressway accident dataset. The analysis reveals that weekdays are the peak period for traffic accidents, with the number of incidents significantly increasing during morning and evening rush hours, peaking during the evening rush hour, while reaching the lowest levels from late night to early morning. Additionally, regardless of weather conditions, “slight” accidents dominate in number. Certain weather conditions may lead to a slight increase in the total number of accidents, but the distribution proportion of severity levels remains relatively stable. Subsequently, a random forest model is established to predict the number of casualties, with an accuracy of 0.7388 on this dataset, which demonstrates favorable fitting performance. Furthermore, we employ the Gini index to measure the influence of selected features on the outcome. The results indicate that the number of vehicles involved in an accident is the core factor affecting casualty counts, while road surface conditions and lighting conditions also significantly impact the number of casualties. These findings provide valuable insights for the application prospects of machine learning algorithms in this scenario.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.