ML-SHAP 기반 도로 유지보수 발생 예측 및 영향요인 검토: 제주도 도로 교통 현황을 중심으로 = Prediction of Road Maintenance Intervention and Assessment of Influencing Factors with ML-SHAP: Focusing on Road Traffic Conditions in Jeju

다국어 초록 (Multilingual Abstract)

Pavement performance deterioration is known to result from the combined effects of traffic demand, axle loading, climatic conditions, and coastal environments. A wide range of models has been proposed worldwide to predict pavement deterioration and the timing of maintenance. In recent years, Jeju Province has experienced a sustained increase in traffic volumes driven by annual visits of more than 10 million tourists, which may accelerate the aging of road infrastructure and raise maintenance costs. In Jeju City, pothole-related complaints are occurring at an average of 3,000 to 4,000 per year, and the number of pothole repairs over the past three years has averaged approximately 2,400 per year. To quantitatively analyze future road maintenance needs, it is required to link Jeju’s distinct road-transport environment, characterized by a maritime climate, frequent rainfall, and a high proportion of coastal roads, with maintenance records. Therefore, this study reviews Jeju’s road-transport conditions and constructs an annual, link-level dataset that integrates road, traffic, and weather factors for both national and local roads. Based on the integrated dataset, a machine learning model is developed to predict future maintenance occurrence, and SHAP (SHapley Additive exPlanations) is applied to interpret the relative importance and directional contributions of key factors associated with maintenance events. The results provide that even when direct pavement condition indices (i.e., cracking and rutting) are not readily available, annual traffic, weather, and road characteristics alone can be used to proactively identify links with a high likelihood of maintenance (intervention) in the following year. The findings provide empirical evidence that can support the selection of maintenance targets and annual budget allocation, as well as priority-setting linked to peak-season traffic management in Jeju.