Road Accident Severity Prediction in India using Machine Learning

India considers road safety to be a serious problem ranking one of the worst in the world. Heavy traffic volume combined with dangerous road conditions and unreliable weather creates unsafe conditions that result in thousands of accidents every year causing unnecessary fatalities and injuries. Predicting the nature of the accident severity on road traffic accidents is difficult. Understanding how serious an accident will be does help emergency responders better prepare and allocate resources, allowing them to proactively respond before an avoidable tragedy occurs. This project’s goal was to create a system based on machine learning to predict levels of severity of accidents related to road traffic using real data from 2017-2022 data. We tested and compared common algorithms (Logistic Regression, Decision Trees, Random Forest, SVMs, and XGBoost) to predict the outcomes of accident severity and found that XGBoost had the most accurate results on accident severity of 91.78%. The machine learning system we developed is trained to classify accidents use the level of severity as ‘Minor,’ ‘Serious,’ and ‘Fatal,’ through learning the underlying complexity incorporated in the road traffic accident data.