Slope instability is the result of a variety of factors, so conventional mathematical models are difficult to get accurate prediction results. To improve the accuracy of slope stability prediction, the improved sparrow search algorithm (ISSA) was used to optimize the support vector machine (SVM). Thus, a new slope stability prediction (ISSA-SVM) model was established. The gravity, cohesion, internal friction angle, slope angle, slope height and pore pressure ratio were treated as input features, and the slope stability state was used as the output. Then, the prediction results of slope stability can be gotten. The domestic and foreign engineering examples were selected to establish a slope database, the ISSA-SVM model was compared with the SSA-SVM model, and the sensitivity analysis was performed using grey relation analysis (GRA). The results show that: the prediction accuracy of the ISSA-SVM model is higher, the generalization ability of the ISSA-SVM mode is stronger, and cohesion and internal friction angle are the most sensitive factors to slope stability. The proposed ISSA-SVM model can not only accurately predict the slope stability state, but also provide reference for the related problems in other areas.