The construction of TBM tunnels is associated with high uncertainty due to the unknown ground conditions surrounding the TBM. Recently, there have been several attempts to make use of the large amount of TBM data recorded during construction to predict the ground conditions and automate the tunneling process. This study presents an implementation of supervised learning models to the Porto metro dataset (Sousa and Einstein 2012) and showcases an alternative method of predicting the ground class. The results of several machine learning (ML) models are reported and compared to each other. These ML models use the same algorithm but with different sets of input features (ie, TBM parameters) to investigate the effect of different TBM parameters on predicting the geology of the tunnel. The results show that the learned model achieved high accuracy when predicting ground classes. Also, it indicates that input feature selection process is a crucial step to build a robust model since it eliminates ambiguous data thus increasing modeling accuracy while reducing training time. Moreover, the confusion matrices of different models showed that the rock ground class had higher scores and consistency under different sets of TBM features. This suggests that other ground classes need more refinement to attain a better model performance.