| dc.description.abstract | The study emphasizes the potential of data-driven methodologies to enhance infrastructure management, promoting sustainability in pavement maintenance and decision-making. Precise pavement structural evaluation is essential for effective management, yet nondestructive techniques like the falling weight deflectometer (FWD) are challenging for extensive networks due to resource and expertise limitations. This study introduces a machine-learning (ML) approach employing linear regression (LR), decision trees (DT), random forest (RF), and gradient boosted trees (GBT) to predict deflection basin parameters (DBP), namely, surface curvature index (SCI), base curvature index (BCI), base damage index (BDI), area under pavement profile (AUPP), deflection ratio (DR), and shape factors (SF) SF1 and SF2. Our methodology leverages the gain ratio criterion for robust feature selection and integrates diverse input variables, including structural, functional, environmental, and subgrade properties. The algorithm, trained and validated in R-software with field trial data, is evaluated using root-mean square error (RMSE), mean square error (MSE), mean absolute error (MAE), and correlation coefficient. GBT consistently outperformed DT and RF, showing lower errors across all parameters: SCI (0.015), BCI (0.085), BDI (0.071), AUPP (0.013), DR (0), SF1 (0.047), and SF2 (0.309). This study underscores the potential of data-driven methods to improve infrastructure management and sustainability in pavement maintenance. | |
