<![CDATA[This study examines factors influencing speeding violations among long-haul freight truck drivers in South Sulawesi using the Random Forest (RF) algorithm. Data were collected from 370 drivers at the UPPKB Datae weigh station in Sidrap Regency, covering variables such as delivery pressure, sleep duration, truck age and size, monthly income, driving experience, daily driving hours, and driver age. Two modeling scenarios were tested: without balancing and with data balancing to address class imbalance. The unbalanced model achieved the highest performance (accuracy = 0.9929; F1-score = 0.9752; AUROC = 0.982), while the balanced model improved minority class detection despite a lower AUROC (0.711). Feature importance analysis revealed sleep duration, truck size, delivery pressure, and driver age as dominant predictors. Biological factors, vehicle characteristics, and operational pressures significantly affected speeding behavior. Policy implications include enforcing limits on working hours, ensuring minimum rest periods, revising incentive structures, providing regular safety training, and employing monitoring technologies such as e-logbooks and GPS. Future research should explore alternative algorithms, advanced balancing techniques, and integration of real-time operational and behavioral data to enhance predictive accuracy.]]>
Copyrights © 2026
