Flow-like landslides are characterized by the long run-out distance, high mobility, and huge impact force. Additionally, the geotechnical parameters of landslide materials exhibit significant uncertainty, leading to inefficiency and prolonged time consumption in analyzing the dynamic processes, predicting influence range, and assessing hazards of flow-like landslides. To address these issues, this study developed a stochastic analysis method for predicting the run-out distance and deposition area of flow-like landslides by integrating artificial neural networks (ANN) and Monte-Carlo simulation techniques with a smoothed particle hydrodynamics (SPH) numerical model. This approach significantly improved the efficiency of predicting the landslide hazard. The accuracy of the SPH numerical model was validated by comparing simulated results with the actual three-dimensional deposition topography of the catastrophic Yigong landslide. Further, the precision of the stochastic analysis method for landslide impact assessment was checked by comparing cumulative probability curves derived from ANN predictions with those obtained from Monte-Carlo simulations combined. Subsequently, by adjusting the mean value, standard deviation, and correlation coefficient of internal friction angle and cohesion of the landslide mass, the intelligent model predicted the frequency distribution and cumulative probability curves of landslide coverage areas, revealing the influence pattern and degree of strength parameters on the probability distribution of landslide coverage area. Finally, a comprehensive hazard assessment index and methodology were proposed, incorporating the failure probability, the unstable slope volume, and the run-out distance. The study also explored the variation in assessment indices under the influence of strength parameter uncertainties and topographic factors for a landslide case. The results demonstrate that the ANN model combined with the SPH numerical model exhibits high efficiency, while the comprehensive hazard assessment index offers a novel approach for the landslide hazard assessment.

flow-like landslide,influence range,artificial neural networks,hazard assessment