Estimation of Unit Maintenance Costs by Tunnel Construction Methods Using XGBoost

The accelerated deterioration of tunnel infrastructure has highlighted the growing importance of effective maintenance.Predicting maintenance costs is a critical component of budget allocation and planning.This study addresses the limitations of conventional methods based on approximate unit costs by proposing a machine learning model that uses actual repair records.A dataset of 6,611 records covering 36 repair methods was collected between 2019 and 2023.The XGBoost algorithm was applied to develop models for cost estimation.As a result, models were developed for 13 methods, achieving an average accuracy of 95.1% and a weighted F1-score of 0.945.The models also achieved an average MAE of 22,822 KRW and an average MAPE of 35.5%.These results show that the proposed approach significantly outperforms traditional unit-price methods.The findings demonstrate the potential of machine learning to improve the accuracy of cost estimation and to support more rational and sustainable budget planning for tunnel maintenance.