2022 Volume 3 Issue J2 Pages 826-847
The purpose of this study is to construct a model that can predict the occurrence of deformations in subway tunnels and determine whether or not percussion inspections should be conducted based on the predictions. The XGBoost model, which has been widely used in recent years, is used for this purpose. However, XGBoost has many hyper-parameters, which require time and effort for parameter tuning. Therefore, we categorized the hyperparameters of XGBoost based on mathematical considerations, conducted a grid search using three selected hyperparameters, and evaluated them based on the average value of AUROC. Finally, some properties of the interrelationships among the hyperparameters λ and γ are proved. As a result, we were able to provide some theoretical basis for the output results. Therefore, we report the results here.
