COMPARATIVE ANALYSIS OF SUPERVISED MACHINE LEARNING ALGORITHMS FOR TRAFFIC ACCIDENT PREDICTION IN OGUN STATE, NIGERIA

Ogun State's frequent road traffic accidents kill, injure, and set back Nigeria's progress. Accident analysis in Nigeria primarily uses descriptive and retrospective methodologies. This hinders their ability to generate accurate predictions and to discern intricate, nonlinear interactions among many elements. This study performed a comparative examination of supervised machine learning techniques to forecast traffic events in Ogun State, Nigeria. Road traffic collisions were recorded quarterly from 2021 to 2024 based on official traffic data. The dataset included information on the overall number of accidents, injuries, fatalities, total cars involved, instances of speeding, driver weariness, driving under the influence, adverse weather conditions, and the time of day. We utilized the orange data mining toolbox to implement four supervised machine learning techniques after data purification, preprocessing, and feature selection. The implemented algorithms comprised K Nearest Neighbors, Decision Tree, Random Forest, and Support Vector Machine. The model's performance was evaluated using accuracy, precision, recall, F-score, specificity, and the area under the ROC curve. The data clearly indicated that the algorithms produced significantly divergent outcomes. The Random Forest model consistently outperformed all other models in every examination. It demonstrated exceptional precision, resilience, and capacity for differentiation. Support Vector Machines excelled in generalization, whilst Decision Trees were adequate yet easily interpretable. The selection of parameters and feature scaling significantly impacted the results, as seen by the suboptimal performance of the K Nearest Neighbors technique. Crash frequency, the number of cars involved, and behavioral characteristics such as acceleration were major drivers of accident outcomes, as demonstrated by the correlation study. The research indicates that ensemble-based supervised machine learning models, especially Random Forest, are the most successful for forecasting traffic occurrences in Ogun State. Data-driven predictive analytics in Nigeria can improve proactive road safety planning, focused enforcement, and evidence-based policy design to reduce traffic accidents.