Predicting Road Traffic Accident Severity Using Machine Learning Algorithms in Kano State, Nigeria

Road traffic accidents (RTAs) are a major global public health and safety crisis, leading to millions of deaths and injuries annually. This study focuses on predicting the severity of RTAs (fatal, serious, or minor) in Kano State, Nigeria, by addressing a key gap in prior research: the limited consideration of all three primary contributing factors (human, environmental, and vehicle) simultaneously. This research utilizes a dataset of 1,368 RTA instances with 19 attributes from the Federal Road Safety Corps (FRSC), Kano State command, covering the years 2019 to 2021. After pre-processing, including handling missing values and applying the Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance, we developed and evaluated five machine learning classifiers: Decision Tree (DT), K-Nearest Neighbor (KNN), Naïve Bayes (NB), Neural Network (NN), and Random Forest (RF). The models were implemented using Python and Scikit-learn. Experimental results demonstrate that the Random Forest algorithm achieved the highest prediction accuracy of 97%, significantly outperforming other methods (KNN: 91%, DT: 91%, NN: 70%, NB: 58%). This high-accuracy model provides a robust framework for road safety authorities to predict accident severity based on causative factors, enabling more targeted and proactive interventions to reduce fatalities.