Evaluating seismic liquefaction potential using shear wave velocity using machine learning

Liquefaction investigations use a variety of approaches based on field and laboratory tests.The paper describes a study designed to determine the probability of soil liquefaction in a region covering Turkey and Iraq.We used machine learning approaches, particularly Random Forest (RF) models, to build and test models to estimate the chance of liquefaction, with shear wave velocity playing a critical role.In addition, earthquake magnitude and peak acceleration were considered important variables.The dataset includes soil attributes such as effective vertical stress (σ′v0), soil type, shear wave velocity (Vs), and earthquake parameters including peak horizontal acceleration (PGA) and magnitude (M), allowing for the computation of liquefaction risk as actual values.The results indicate that Random Forest predicted soil liquefaction potential with a remarkable 92.5% accuracy using only 20% of the dataset.This study adds to the progress of risk assessment approaches in earthquake-prone locations, hence improving infrastructure resilience and catastrophe protection.