Exploring high-order multivariate geotechnical features using the minimum information dependence model

Understanding the complex interdependencies among soil parameters is crucial for designing and maintaining civil engineering structures. Traditional modeling often assumes linearity and normality, potentially overlooking the nonlinear and higher-order relationships inherent in geotechnical data. This study investigated the application of a new statistical model, the Minimum Information Dependence Model (MIDM), to extract new features from multivariate soil parameters by capturing higher-order dependencies. First, we reviewed the theoretical foundation of MIDM and developed algorithms tailored for geotechnical applications. To evaluate the effectiveness of the model, we conducted two case studies: a synthetic dataset with intentionally embedded higher-order dependencies and a real-world dataset from the Tokyo-CLAY14/67760 geotechnical database comprising two distinct types of soft clay. Our results demonstrate that MIDM successfully identifies complex nonlinear interactions among soil parameters that traditional linear models fail to capture. In the synthetic case, MIDM accurately recovered higher-order dependencies and provided reliable posterior distributions. In the real-world case, the model revealed subtle differences in soil properties associated with different classifications, enhancing the predictive accuracy of geotechnical parameter estimation. We conclude that MIDM offers significant potential for advancing geotechnical engineering analysis by providing a unified framework capable of handling both continuous and categorical data.