As coal mines and underground engineering projects extend into deep and geologically complex regions, the application of Tunnel Boring Machine (TBM) in rapid roadway excavation has become increasingly widespread. However, highly variable and complex geological conditions often lead to severe equipment load fluctuations, accelerated cutter wear, and reduced excavation efficiency, which significantly hinder the stable and efficient operation of TBM. Therefore, it is essential to establish an accurate boreability prediction and rock mass classification method to provide a scientific basis for TBM parameter optimization and construction planning. This study proposes an intelligent prediction model—GWO-VMD-SSA-LSTM—by integrating Grey Wolf Optimization (GWO), Variational Mode Decomposition (VMD), Sparrow Search Algorithm (SSA), and Long Short-Term Memory (LSTM) networks. The model enables accurate prediction of TBM boreability and quantitative classification of surrounding rock. Results show that the model achieves excellent performance on the test set, with a MAE of 0.4324, RMSE of 0.6005, MAPE of 1.5486%, and R2 of 0.9527, significantly outperforming other comparison models in terms of accuracy and generalization ability. Furthermore, a rock mass classification system based on the Field Penetration Index (FPI) was developed, enabling rapid determination of boreability levels. Engineering validation demonstrates that the method effectively reflects excavation difficulty across various lithologies and provides intelligent decision-making support for TBM tunneling in complex geological conditions, offering significant practical value for advancing intelligent excavation in coal mine roadways.