Novel multi-spatial receptive field (MSRF) XGBoost method for predicting geological cross-section based on sparse borehole data

Due to the complex spatial features of geological formation, it remains a significant challenge to accurately predict geological cross-sections from limited borehole data. This study develops an innovative multi-spatial receptive field (MSRF) XGBoost approach, which encompasses classification and identification modules to forecast geological cross-sections using sparse borehole data. The classification module exclusively employs sparse borehole data to train a series of MSRF XGBoost models for soil classification. The identification module leverages all the trained models to generate potential predictions of unknown soil strata, automatically pinpointing the optimal one via Gaussian filtering and boundary similarity algorithms. A new boundary accuracy criterion is proposed to assess the prediction capacity of different models. Following this, the developed MSRF XGBoost method is compared with an existing conventional XGBoost method using both linear and nonlinear cases. The findings illustrate that our proposed method enhances the prediction accuracy for both linear and nonlinear geological cross-sections. Furthermore, the developed method is employed to determine a geological cross-section in the Netherlands using open-source borehole data. The accuracy of the method in predicting soil layers in all in situ boreholes reaches an impressive 90%, validating its effectiveness in practical geotechnical engineering.