The global sensitivity analysis of slope stability based on the least angle regression

The least angle sensitivity (LARS) algorithm is used to realize the global sensitivity analysis of slope stability, while simultaneously considering the effects of several geotechnical parameters on slope stability. In addition, the Sobol sequence is applied in the sample simulation to generate the geotechnical parameters, thereby increasing the accuracy of the results. Two cases are considered to investigate the effects of the geotechnical parameters on slope stability, and the accuracy and efficiency of the LARS algorithm are examined. The importance measure indexes obtained using the LARS algorithm are in good agreement with those obtained using the Monte Carlo (MC) method. To determine the importance measure indexes, the performance functions of the slope stability analysis are required to be run N times when using the LARS algorithm, which is 1/(n · N + 1) the required number for the MC method, where n and N represent the number of random variables and sample size, respectively. In this scenario, the computational efficiency is considerably increased.