|
|
 |
| |
|
|
|
Predicting and Analysis of Traffic Accident Severity and Risk Factors in Jordan by using Machine Learning |
|
|
|
PP: 1419-1426 |
|
|
|
doi:10.18576/amis/190615
|
|
|
|
Author(s) |
|
|
|
Yahia Ali Saad Khalayleh,
Suleiman Ibrahim Mohammad,
Nawal Salem Al Louzi,
Mahmoud Baniata,
Mohammad Hassan,
Mohammed Abu Safaqah,
Asokan Vasudevan,
Muhammad Turki Alshurideh,
|
|
|
|
Abstract |
|
|
| In this study, machine learning techniques are applied to the JO-Traffic-Accidents-Dataset (JO-TAD) to investigate traffic accidents in Jordan, which contains 73,095 traffic accident reports for 2018. Based on the three machine learning models Random Forest, XGBoost, and Neural Network, we predicted the accident severity and highlighted major factors contributing to accident severity. The best-performing model was XGBoost with 87.1% accuracy, followed closely by Neural Network (86.2%) and Random Forest (85.3%). Weather conditions were found to be the most significant factor (importance score: 0.85), followed by road type (0.79) and driver age (0.73). It was found that the winter months witnessed a greater severity of accidents, with a 23% higher rate of severe accidents than in summer months. This data-driven study could benefit to conduct purposeful policy succession and management in traffic safety in Jordan. |
|
|
|
|
|
|
|
