Abstract: The landslide susceptibility evaluation is the basic work of landslide management, and it is also an important basis for formulating various disaster prevention and mitigation measures. In view of the low accuracy of the traditional information model in determining the weight value in the evaluation process, this paper proposes a coupling model of RBF neural network and Information value model. 9 index factors such as slope are selected to build the evaluation index system of landslide susceptibility in Min Xian of Gansu Province. The RBF neural network information value coupling model (RBFNN-I) is used to carry out the landslide hazard susceptibility evaluation. Rationality test and ROC curve are used to test the accuracy of the evaluation results of the model. The results show that: (1) the AUC value of RBFNN-I model is 0.853, which is 6.3% and 9.7% higher than that of single RBFNN and I model, respectively, indicating that RBFNN-I model has better evaluation accuracy; (2) the extremely high and high susceptible areas of landslide disasters in Min Xian are mainly distributed along Lintan-Dangchang fault zone, Tao He and its tributaries, and the valleys on both sides of Lyuning River and Puma River.