Pufferfish-based ensemble models to estimate the stiffness modulus of soft ground soil

Soft soils present significant challenges to environmental and infrastructure development due to their unique characteristics, which include low bearing strength, high water content, restricted permeability, and a high void ratio. One of the most crucial considerations while constructing geo-structures is the stiffness modulus (Gs) of soft ground soils. Determining the stiffness modulus of soft ground materials, including soils, requires expensive equipment, additional expert labor, and time. Conversely, current trends in sustainable development support the use of less expensive technology. AdaBoost Regressor (ABR) and Extra Tree Regressor (ETR) were used in the computation. The ABR and ETR hyperparameters, which need to be selected using metaheuristic optimization approaches, have a significant influence on its dependability. Using the Pufferfish algorithm (PA) is how this is achieved. A machine database including 197 examination findings from earlier research indicates that the Gs of soft ground soils is affected by six different input parameters. The data suggests that both ABE(PA) and ETR(PA) have a considerable probability of precisely ascertaining the value of Gs. The ABR(PA) produced the lowest MedAE metric values during both the training and evaluation phases, with values of 0.819 and 0.9685, respectively. Conversely, the ETR(PA) exhibited a lower level of dependability throughout both the training and evaluation phases, with values of 0.727 and 1.184, respectively. Received: 13.03.2025 Received in revised form: 21.01.2026 Accepted: 12.06.2026