Prediction of Pedestrian Traffic Accident Severity and Evaluation of Driver and Pedestrian Faults

This study investigated pedestrians involved traffic accidents with the aim of classifying the severity of accidents based on the number of injured pedestrians using machine learning algorithms, including AdaBoost, Gradient Boosting, XGBoost, K-Nearest Neighbors (KNN), Logistic Regression, Support Vector Classifier, Decision Tree, and Random Forest. The Random Forest model was identified as the best model for classifying pedestrian-involved traffic accidents, achieving high predictive accuracy of 95%, an F1 score of 0.95, and demonstrating low error metrics. The research analyzed both driver and pedestrian faults, alongside factors such as the presence of pedestrian crossings, intersection type, driver age, time of day, month and seasonal variations. The results revealed that accidents at locations without intersections were primarily caused by driver faults, such as speeding, while pedestrian faults, such as crossing at unintended locations, also significantly contribute to the overall accident rate. The findings offered valuable insights into the characteristics of pedestrian accidents to improve traffic safety and reduce pedestrian injuries and fatalities.