Research on Geotechnical Data Interpolation and Prediction Techniques

The development of underground space is vital for urbanization and infrastructure projects. Prior to construction, comprehensive geological exploration is essential to ensure stability and safety. However, acquiring complete and accurate statistical data for project management is challenging, necessitating the handling of missing data to enhance reliability. Interpolation techniques are an effective way of dealing with incomplete data. This study presents a scalable framework for geotechnical data interpolation using machine learning. The framework employs different regression models to construct estimators and accurately interpolate geotechnical data. Key considerations include model selection and parameter optimization, with complete data used as the regression target. Five regression models, Bayesian Ridge Regression (BR), Extreme Gradient Boosting Tree (XGBoost), Support Vector Machine (SVR), Random Forest (RF) and K-Nearest Neighbour (KNN), were utilised. Estimators are constructed using the regression models and iterative interpolation is used to estimate missing values for geotechnical data, with each feature treated as a result of using the different estimators. The framework is evaluated through k-fold cross-validation, demonstrating its effectiveness in imputing missing values. The interpolation results using the SVR model indicate good conformity with the original data, confirming the method's effectiveness in capturing underlying patterns. This scalable framework bridges the gap in geotechnical data interpolation research, providing a reliable solution. The proposed approach contributes to the accurate and robust interpolation of geotechnical data, facilitating informed decision-making in underground construction projects.