Application of machine learning and big data in enhancing road safety by predicting accident severity

Abstract

Background: Road traffic accidents have been a recurring global problem that cause over 1.3 million lives annually, with a significant impact on the mortality of children and young people. Effective accident prevention strategies have been required due to the ensuing expenses to society, the legal system, and the hospital system. The goal of the current work is to create a safe system that integrates engineering, traffic control, and vehicle standards for accident predictions to improve road safety. Method: The algorithms Gradient Boosting Classifier, XGB Classifier, and Random Forest Classifier were utilized to forecast accident severity and produce an interactive map that could locate accident hotspots and estimate accident frequencies. Results: In total 660,679 cases the dataset provided a solid basis for examining trends and patterns in traffic safety with the Gradient Boosting Classifier accurately predicting the model with over 85%. A total of 16.74% were classified as serious accidents and 83.26% as slight. In the validation phase, our final accident prediction model, utilizing Gradient Boosting Classifier algorithm, showcased outstanding performance metrics. With an accuracy score of 0.85, it demonstrated exceptional proficiency and was highly likely to be accurate when predicting a positive instance. However, both its precision stood at 0.71, indicating that while it was accurate, it was still effective at detecting all positive instances or achieving total correctness. Conclusion: The accident severity prediction model demonstrated accuracy and reliability, as evidenced by its successful external validation using high-quality registry data within the UK. This model holds significant potential for deployment in improving road safety analytics, enabling stakeholders to effectively assess and compare accident severity outcomes across different regions and transportation systems. KEY WORDS: Road Safety, Machine learning, algorithms, Big Data, Accident Prediction, Road Safety, Classification Techniques.