Traffic Accident Severity Classification System Using Random Forest Algorithm

Abstract

Traffic accidents pose a major concern in many countries, including Indonesia, causing considerable losses, injuries, and fatalities each year. Properly classifying the severity of these incidents is essential for authorities to establish preventive actions, apply effective countermeasures, and improve overall road safety. Conventional statistical techniques often fall short in capturing the intricate relationships among multiple influencing variables, such as weather, driver experience, vehicle type, number of vehicles, and casualty figures. To address this limitation, this study proposes a machine learning–based classification method using the Random Forest algorithm, known for its robustness in handling complex and high-dimensional data while identifying nonlinear patterns. The model was trained on a traffic accident dataset from Kaggle and incorporated important features, including driver age group, driving experience, type of vehicle, lighting and weather conditions, type of collision, number of vehicles involved, and casualties. The proposed system achieved 81% accuracy, 75% weighted precision, 81% weighted recall, and a weighted F1-score of 77%, demonstrating reliable performance in predicting accident severity levels Slight Injury, Serious Injury, and Fatal Injury. And providing useful insights for data-driven planning in traffic safety management.Traffic accidents pose a major concern in many countries, including Indonesia, causing considerable losses, injuries, and fatalities each year. Properly classifying the severity of these incidents is essential for authorities to establish preventive actions, apply effective countermeasures, and improve overall road safety. Conventional statistical techniques often fall short in capturing the intricate relationships among multiple influencing variables, such as weather, driver experience, vehicle type, number of vehicles, and casualty figures. To address this limitation, this study proposes a machine learning–based classification method using the Random Forest algorithm, known for its robustness in handling complex and high-dimensional data while identifying nonlinear patterns. The model was trained on a traffic accident dataset from Kaggle and incorporated important features, including driver age group, driving experience, type of vehicle, lighting and weather conditions, type of collision, number of vehicles involved, and casualties. The proposed system achieved 81% accuracy, 75% weighted precision, 81% weighted recall, and a weighted F1-score of 77%, demonstrating reliable performance in predicting accident severity levels Slight Injury, Serious Injury, and Fatal Injury. And providing useful insights for data-driven planning in traffic safety management.