In order to solve the problem that the geological condition in front of the cutterhead is not clear in the shield construction process, this study proposes a stratigraphic classification model using machine leaning technology based on the relationship between shield parameters and geological characteristics. Shield parameters are dynamic response of geological characteristics during shield tunnelling. The transformed shield parameters, specific thrust and specific torque are used to reveal the changes of geological characteristics. CART model and random forest model are developed to predict stratigraphic categories during shield advancing. The operation parameters of shield in Pazhou railway project are collected and fed to machine learning-based model for training and prediction. The results show that the accuracy of CART prediction model and random forest model are 81% and 85%, respectively. There is a small error on the boundary of local adjacent categories, which meets the actual engineering requirements and can provide reference for the construction of similar projects.