In drill-and-blast tunnels, the recorded data from fully computerized rock drilling jumbos can reflect the tunnel face rock mass conditions. An intelligent method based on real-time drilling data was developed herein to identify tunnel rock mass grades. First, the time series of 14 drilling parameters was processed using a Fourier transform, and the maximum amplitudes of each parameter were computed. Subsequently, the means, medians, and variances of the parameter amplitudes from different blast holes on the same tunnel face were satisfied to form the original feature set. Neighborhood component analysis was used to select the key rock mass condition features. From this feature selection (FS), the top five features were selected as the final feature set. Furthermore, using the selected features, a Gaussian Process Classification (GPC)-based model was established to predict rock mass grades. The proposed method captures both the drilling data information in the time dimension and the inherent connection of different blast hole data, making it advantageous. Using field construction data from Xiwuling Tunnel, the proposed FS-GPC model performance was evaluated. It achieved an accuracy of 90% in predicting rock mass grades. The proposed method can accurately and rapidly predict rock mass grades in drill-and-blast tunnels, optimizing engineering design and construction decisions.