ABSTRACT
Predicting the input design parameters of clay soils accurately is essential for the effective design of flexible pavements, as it serves as a critical indicator of the strength and stability of soil subgrades. Traditional laboratory methods for determining CBR can be labour-in-tensive and time-consuming. This study addresses the pressing challenge by implementing opti-mized ML techniques to predict CBR, using six fundamental geotechnical parameters: OMC, MDD, PL, FC, PI, and LL. Three ML models—Light GBM, SVM and random forest—were employed and fine-tuned through grid search optimization to ensure superior predictive perfor-mance. The model’s effectiveness was evaluated using metrics RMSE, R2 and MAE. The pro-posed ML models offer a rapid, reliable and cost-effective alternative to traditional CBR testing, by significantly minimizing the resources and time needed for subgrade evaluation. By improving the pavement design process, these models contribute to the advancement of geotechnical engi-neering and the development of modern infrastructure.
