Application of XGBoost in Road Maintenance Cost Prediction

Road maintenance costs play a critical role in government budgeting, as they represent a recurring expenditure required to sustain transportation infrastructure performance and traffic safety. Accurate cost prediction enables long-term efficiency by ensuring that maintenance budgets are allocated appropriately. This study aims to develop a predictive model for road maintenance cost using the Extreme Gradient Boosting (XGBoost) algorithm, optimized through iterative training to improve prediction accuracy based on deviations between predicted and actual costs. Model performance was evaluated using Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R²), all of which indicate a strong model fit and high predictive reliability. The model was developed using simulated and empirical data from 30 road sections with varying characteristics, incorporating key predictors such as road length, cold mix asphalt, asphalt emulsion, diesel fuel, gasoline, water consumption, working area, asphalt removal volume, and labor requirements. The results demonstrate that the proposed XGBoost-based model can effectively estimate maintenance costs and associated resource requirements. The findings provide practical insights for government agencies in planning material usage and workforce allocation for road maintenance activities.