Landslide susceptibility assessment in Qingyijiang river basin based on different machine learning methods

Susceptibility assessment is an effective tool for landslide disaster prevention and control. Machine learning algorithms can effectively reduce the subjectivity of determining the weights of factors in disaster risk assessment, which has become a hot spot in current research. We selected maximum monthly rainfall during the year, slope, FVC, distance from rivers, distance from faults and vegetation type as the influencing factors for landslide susceptibility assessment, and used BP neural network and random forest algorithms to evaluate the susceptibility of landslide hazards in the Qingyijiang river basin, and compared the evaluation results. The results show that the fitting results of the random forest algorithm are better. According to the fitting results of the random forest algorithm, the landslide hazard in the Qingyijiang river basin is closely related to the distance from the fault and the distance from the river, and about half of the area (48.75%) in the basin is at risk of small landslides, and most of them are distributed near the river.