In order to predict the slope stability, a method based on machine learning is put forward. By analyzing the slope instability characteristics, six typical slope parameters (pore water pressure, slope height) are selected and analyzed, and the corresponding evaluation data set is established. The slope stability model is established by using machine learning methods such as gradient hoist (GBM) and support vector machine (SVM), and verified by 50% cross-validation. By analyzing the characteristics of slope instability and combining with domestic slope cases, according to the sensitivity of characteristic parameters, the slope protection measures for different sensitive factors are put forward.