Application of Random Forest and SMOTE in Predicting Fatal Traffic Accidents: The Case of Erzurum

<jats:p xml:lang="en">This study presents a machine learning model developed using the Random Forest algorithm to predict whether traffic accidents in Erzurum will be fatal. Data from 16793 traffic accidents that occurred between 2014 and 2023, provided by the General Directorate of Security, was used. This dataset includes various variables such as driver characteristics, weather conditions, road type, road condition, lighting, shoulder, etc. Due to the minority of fatal accidents in the dataset, class imbalance was addressed using the SMOTE (Synthetic Minority Over-sampling Technique) method. The model was tested on training and test data with high performance metrics such as 98% accuracy, sensitivity, and F1 score. The results obtained reveal the impact of variables such as accident type, driver age, and number of vehicles on fatal accidents, contributing to data-driven policy development processes aimed at improving traffic safety.</jats:p>