Abstract: Rockburst disasters constantly occur in the construction of mines, tunnels(caves),foundation pits, and other underground projects at home and abroad, and their frequency and intensity increase with the extension of projects into deeper areas. Therefore, predicting the intensity of rockburst levels has become an important research direction. By selecting maximum shear stress(σ_θ),uniaxial compressive strength(σ_c),uniaxial tensile strength(σ_t), and elastic energy index(W_et) as four characteristic values of the prediction model, this paper presents a combined model based on the SMOTEENN-CGAN data processing integrated algorithm that is used to predict the rockburst intensity level. In this model, SMOTEENN and CGAN algorithms are used to process the original data through the methods of over-sampling, under-sampling, and augmented generation. Then, ten classical algorithms are used to verify the validity of SMOTEENN-CGAN. Finally, four groups of rockburst intensity level prediction models with different base models and meta-models are constructed with the Stacking integrated algorithm. The results indicate that:(1)SMOTEENN-CGAN can be used to deal with multiple classification problems, and the newly generated rockburst data is consistent with the original distribution characteristics, and the dispersion degree of pre-processed data and abnormal points are significantly reduced; (2)After data preprocessing, the performance of ten classical algorithms has been improved to varying degrees, and the average accuracy of each algorithm has increased by 1.87% to 7.75%; MLP and NP increased by 7.75% and 7.43%,respectively. (3)Different combinations of base models and meta-models affect the performance of Stacking. In combination ④, when the base model is XGBoost+LGBM+ETC, the highest prediction accuracy rate of Adaboost in the meta-model is 96.12%. When the reliability of the rockburst intensity level prediction model is verified by engineering examples, the maximum prediction accuracy can reach 92.3%. The model in this paper provides an effective and feasible machine learning prediction method for rockburst intensity prediction.