Traffic accidents in Surakarta City, Indonesia, present a significant challenge, necessitating effective predictive models to mitigate their impact. This study examines the use of the Decision Tree (DT) algorithm to predict the severity of road traffic crashes and identify key contributing factors. Utilizing traffic accident data from the Surakarta Police Department covering 2018–2022, the study analyzes variables such as collision type, road function, and road class. The Kendall correlation matrix identified strong relationships between these variables, particularly between road function and class categories. The DT model revealed that collision type 6 (self-collision) and road class III are the most influential factors. While the model demonstrated strong classification performance for class 1 (light accidents), it highlighted a need for improved classification of other severity levels. Feature importance rankings further emphasized the critical role of collision types and road categories in predicting accident severity. These findings underscore the potential of DT algorithms in enhancing road safety measures and provide actionable insights for policymakers. Future research should address class imbalances. Integrating ensemble learning techniques and advanced machine learning methods could enhance model performance across all severity levels.