Database Development and Repair Cost Prediction Based on Infrastructure Maintenance Records

Abstract: This study addresses the need for precise maintenance cost prediction by developing unit cost models based on element-level repair data. Using infrastructure maintenance regulations in Korea, a systematic facility breakdown structure was established, and a cost database was compiled from maintenance records. The Extreme Gradient Boosting (XGBoost) algorithm was applied to develop cost prediction models for different repair methods in structural elements such as decks, piers, and drainage systems. The results showed high predictive accuracy, with bridge models averaging 87.2% and tunnel models 85.5%. The findings provide data-driven insights to enhance maintenance planning and support cost-efficient infrastructure management strategies.