Optimizing blasting plans according to actual rock mass classification represents a critical approach for enhancing blasting outcomes. However, field construction conditions present significant challenges in obtaining direct rock mass classification data across different excavation face zones. This study proposes a rock mass classification method utilizing on-site drilling parameters, with application to a Chongqing tunnel project. Firstly, the characteristics of the rock mass were investigated, and the rock mass class was dⅣided. The drilling data were then collected, and drilling parameters related to the rock mass class were screened. Furthermore, the data volume of small sample classes was expanded by the Mahalanobis Distance-based Over-sampling technique (MDO). Meanwhile, the relationship between drilling parameters and rock mass class was modeled using the Extreme Gradient Boosting (XGBoost) algorithm, and an identification model for rock mass class was then established. Finally, the site blasting scheme was optimized based on the identification results. The results show that the MDO-XGBoost model achieves an overall classification accuracy of 80% for rock mass grade identification. The optimized blasting scheme has increased blasting penetration and shortened deslagging time, based on the rock mass classification results. This research presents a viable method for accurately identifying rock mass classes based on drilling parameters, particularly in the context of sample imbalance, thereby contributing to intelligent blasting and efficient construction in tunnel projects.