Multivariate adaptive regression splines model for prediction of the liquefaction-induced settlement of shallow foundations

Buildings with shallow foundation are vulnerable to liquefaction-induced settlement during an earthquake. Accurate settlement evaluation is an important step in earthquake damage mitigation. This paper presents a simplified approach to estimate the liquefaction-induced settlement of buildings with shallow foundations. The multivariate adaptive regression splines (MARS) algorithm is adopted. The validated finite difference method is used to produce artificial data that consider various properties of the soils, structures and ground motions. A relative importance analysis is conducted to quantify the effect of each input parameter and their coupled interactions on the liquefaction-induced settlement. The accuracy of the established model is demonstrated through centrifuge test results available in the literature. •A MARS model is developed to predict the liquefaction-induced settlement of shallow foundations.•The relationship between a series of affecting factors and the liquefaction-induced settlements is described.•The influences of affecting factors on the liquefaction-induced settlements have been quantitatively analysed.•The accuracy and precision of the developed MARS model are validated using collected centrifuge tests in prior literatures.