In order to accurately judge slope stability in a generalized and efficient sense, a new CatBoost ensemble learning algorithm was introduced to evaluate slope stability. According to the basic geometry and geological elements of the slope, without the need for numerical modeling and calculation, the stability state of the slope is directly and objectively evaluated, and the probability of slope stability was given from the perspective of probability information theory, so as to realize the stability evaluation of the slope at the wide-area scale with fast speed, high precision and good robustness. Taking five main factors affecting slope stability as evaluation indexes, a large slope stability evaluation dataset was created, and a slope stability prediction model based on CatBoost ensemble learning was constructed. The simulation results show that compared with common machine learning and representative ensemble learning models, the prediction effect of CatBoost model is significantly better than that of other models, and the hyperparameter tuning requirements are reduced, which is more versatile, objective and reliable, and can be effectively applied to the preliminary evaluation of slope stability. Through the example of a slope project of Renhua to Xinfeng section of Renhua (the junction of Hunan and Guangdong Province) to Boluo Expressway, the feasibility of slope stability evaluation method based on CatBoost ensemble learning is verified.