Abstract: The randomness in geotechnical parameters introduces uncertainty in the dynamic characteristics(velocity, distance, etc.)of landslides. Monte-Carlo simulation(MCS)is an effective approach to account for the influence of geotechnical parameter uncertainty in landslide run-out analysis, but the large number of simulations required for accurate results leads to excessive computational time. Therefore, based on a deterministic analysis method for the process and run-out distance of flow-like landslides, a fast-predicting model of landslide run-out distance was constructed using an artificial neural network(ANN)algorithm, and a stochastic analysis method considering the uncertainty of strength parameters was established by combining Monte-Carlo simulation. Numerical analysis methods were employed to reconstruct the movement process of the Yangbaodi landslide. The simulated run-out distance and deposition morphology were compared with field observations, validating the effectiveness of the SPH numerical model for landslide simulation. Subsequently, variations in soil strength parameters(particularly the internal friction angle)were introduced to obtain run-out distances under different strength conditions, thereby establishing both training and testing datasets for the neural network. Using the trained neural network model, the landslide run-out distance under different strength parameters was predicted, and the performance of the trained model was evaluated by comparing the results with MCS, which demonstrated that the proposed stochastic model can serve as a surrogate for landslide run-out analysis. Finally, a dataset of internal friction angles was generated under a specific probability distribution as input, and the run-out distance of each sample was predicted based on the trained neural network model to discuss the probability distribution characteristics of run-out distance under strength parameter uncertainty. The results showed that when the internal friction angle conformed to a lognormal distribution, the mean run-out distance was approximately 312 m and mainly distributed between 288 m and 324 m, behaving as a random variable with specific probability characteristics. The results demonstrate that the surrogate model developed in this study effectively reduces the computational cost of Monte-Carlo simulations. Furthermore, this study investigates the probability distribution of run-out distances under strength parameter uncertainties, providing valuable references for identifying critical prevention zones in potential landslide hazard areas.