Truck Accident Prediction and Risk Factors Analysis in Jordan: A Machine Learning Approach

In recent decades, Jordan has experienced significant population and urban development growth, which has been associated with a surge in traffic accidents and their negative impacts on individuals, properties, and the local economy. Trucks are more likely to be involved in fatal accidents due to a combination of factors, including driver behavior, traffic conditions, highway characteristics, and environmental conditions. Thus, the factors affecting the accident risk rate among truck drivers in Jordan have been studied, along with the ability to predict the number of accidents and the number of fatalities that the driver could contribute to their career as a truck driver. The results show that the proposed feature selection methodology successfully chooses the significant factors that affect the accident risk rate, indicating that the driver’s behaviors and fatigue-related features affect the type and severity of truck accidents. The results show that using the overbalancing technique (SMOTE) enhances the prediction models' accuracy and decreases the false positives among minor classes as follows: The prediction models can predict the potential of the driver’s involvement in an accident with an accuracy of 84.5% using balanced data compared to 75.4% using imbalanced data; the number of accidents with an accuracy of 85.5% using balanced data compared to 69.7 % using imbalanced data; and the number of fatalities with an accuracy of 85.6% using balanced data compared to 78.5 % using imbalanced data set.