Machine Learning-Based Decision Support Framework for Smart Infrastructure Health Monitoring and Sustainable Civil Engineering Management
Abstract
Good pavement management means good condition assessment and timely maintenance prioritization. The adoption of historical inspection data along with roadway and environmental data can enhance data-driven infrastructure monitoring and aid sustainable asset management. This study developed a machine learning-based decision support framework for estimating pavement condition and translating predicted Pavement Condition Index (PCI) values into maintenance recommendations. Six complementary datasets covering historical and current pavement inspections, flood susceptibility, ground deformation, flood frequency, and topographic characteristics were integrated using a common road-segment identifier. After preprocessing, 11,775 observations and 23 predictors were retained. Linear Regression, Random Forest, and Extreme Gradient Boosting (XGBoost) were trained using an 80:20 train-test split and evaluated through R², RMSE, MAE, observed-versus-predicted analysis, and residual diagnostics. Linear Regression achieved the best performance, with an R² of 0.9885, RMSE of 2.6525, and MAE of 1.5342. Predicted and observed PCI values showed close agreement, while residuals were centered near zero, indicating limited systematic bias. The decision-support stage classified 31.34% of test road segments for routine monitoring, 21.83% for preventive maintenance, 16.39% for minor rehabilitation, 13.59% for major rehabilitation, and 16.86% for immediate reconstruction. The proposed framework provides an accurate and practical approach for pavement condition assessment and maintenance prioritization by integrating pavement, roadway, and environmental information, thereby supporting smart infrastructure health monitoring and sustainable civil engineering management.