A HYBRID ENSEMBLE LEARNING APPROACH FOR PREDICTING ROAD ACCIDENT SEVERITY USING RANDOM FOREST AND BAGGED TREE CLASSIFIER

Abstract

Road accidents are a major public safety concern, leading to injuries, fatalities, and economic losses worldwide. Traditional accident severity analysis methods, such as the Apriori algorithm, have been used to extract association rules among risk factors, but they often lack predictive accuracy and robustness. To overcome these limitations, this study proposes a hybrid machine learning approach that integrates Random Forest for feature selection and a Bagged Tree classifier for accident severity prediction. The proposed system efficiently identifies critical factors influencing accident severity, such as lighting conditions, road surface conditions, weather, and driver behavior. By leveraging ensemble learning techniques, the model enhances prediction accuracy and generalizability compared to conventional statistical approaches. The study also highlights how improvements in road lighting and infrastructure can significantly reduce accident severity. Extensive experimentation is conducted using historical accident datasets, and model performance is evaluated using accuracy, precision, recall, and F1-score. The results demonstrate that the hybrid approach outperforms traditional algorithms in predicting accident severity, making it a valuable tool for policymakers and transportation authorities. The findings provide actionable insights for road safety improvements, accident prevention strategies, and smart traffic management systems, ultimately contributing to safer road environments.

Author

Mrs.M.Sathiyapriya, Mrs.D.Nivethini, Dr.N.Sathyabalaji