An efficient framework of optimized ensemble paradigm for estimating resilient modulus of subgrades

This study employs an efficient framework of ensemble and meta-heuristic optimization algorithm for estimating resilient modulus (M_R) of subgrades with and without considering the influences of freeze–thaw cycles. Notably, M_R is one of the most important stiffness characteristics used in pavement design. The proposed framework combines an ensemble paradigm, random forest regression (RFR), and a widely used meta-heuristic optimization algorithm, grey wolf optimizer (GWO). The outcomes of the established RFR-GWO framework were compared with six regression and neural network-based paradigms namely linear regressor, Gaussian process regression, support vector regressor, artificial neural network, emotional neural network, and multilayer perceptron neural network. For model design and validation, two datasets of A-4, A-6, and A-7–6 (as per AASHTO classification) soils were gathered from the literature. As per experimental results, the developed RFR-GWO achieved the highest degree of accuracy against both datasets with the coefficient of correlation ranging between 0.9970 and 0.9880. To demonstrate the robustness of the established RFR-GWO framework, the impact of the influencing parameters was also investigated via parametric analysis. Overall, the developed RFR-GWO has demonstrated its capability to assist engineers in estimating the subgrade M_R during the initial stage of engineering projects.