The study of the slope stability has the characteristics of non-linearity, complexity and complicated influence factors. In order to find a more accurate evaluation of slope stability, a slope stability evaluation model based on K-means clustering algorithm and neural network is proposed. It is found that the K-means neural network is feasible and accurate in slope stability analysis. By comparing the advantages and disadvantages of K-means clustering on the high efficient inherent hierarchical merging ability and self-learning ability of neural network , 45 groups of experimental data were selected, and 6 groups of influencing factors, which were bulk density, internal friction angle, cohesion, slope angle, slope height, pore pressure ratio, were analyzed and filtered out valid data through the improved K-means algorithm method Then the input data were put into a large number of training and adjustment of weight through the neural network, in order to output the safety factors of stability evaluation. Prediction results show that the predictive ability of the model to the stability of the slope is higher than that of the same type analysis method.