Algorithm-Based Supervised Learning Cost Estimation and Prediction Models for Trenchless Cured-in-Place Pipe Sewer System Rehabilitation

Cured-in-place pipe renewal technology (CIPPRT) has become a preferred trenchless rehabilitation method for aging sewer systems due to its minimal surface disruption, structural reliability, and long-term performance. However, estimating construction costs for CIPPRT projects remains challenging, as traditional empirical or deterministic approaches fail to capture nonlinear interactions among engineering and project variables. This study develops a supervised machine-learning (ML) framework to estimate and predict construction costs for CIPPRT installations across a wide diameter range of 150–2,130 mm (6–84 in.). A comprehensive dataset of more than 500 completed municipal CIPPRT projects executed between 2011 and 2025 across the United States was compiled through a triangulated approach integrating bid tabulations from actual projects, peer-reviewed literature, and published case studies. All costs were standardized to 2025 US dollars and refined through outlier removal, inflation normalization, and multiple imputation by chained equations. Exploratory data analysis and correlation-matrix evaluation revealed strong positive correlations between construction cost and pipe diameter and moderate correlations between liner thickness and cost and between pipe length and installation year, validating the engineering logic embedded in the dataset. Four supervised ML algorithms, multiple linear regression, k-nearest neighbors, decision tree regressor, and gradient boosting regressor, were trained using an 80∶20 stratified split and optimized through grid-based hyperparameter tuning and five-fold cross-validation. Among these, the GBR model achieved the highest predictive accuracy, with a test R2 of 0.80 and a cross-validated R2 of 0.92, along with the lowest root mean squared error of 525.7 USD/m2 (160.22 USD/ft2), mean absolute error of 165.3 USD/m2 (50.40 USD/ft2), and mean absolute percentage error of 15.8%. Its narrow 95% prediction interval [521.3–1,022.6 USD/m (158.88–311.40 USD/ft)] confirmed superior generalization and stability across folds. Model validation revealed a nonlinear increase in cost ranging from 604 to 7,960 USD/m (184 to 2,427 USD/ft), consistent with material, liner thickness, and equipment demand effects. The final GBR-based framework was deployed within a graphical user interface that allows engineers and municipal planners to generate rapid, reliable, and transparent cost predictions for future CIPPRT projects. The integration of cross-validated ML modeling with engineering interpretation establishes a scalable, data-driven foundation for cost forecasting and sustainable decision-making in trenchless sewer rehabilitation.