Regional Landslide Susceptibility Evaluation Based on Random Forest Weighting Information: a Case Study of Zigui to Badong Section in the Three Gorges Reservoir Area

In the study of landslide susceptibility assessment, conventional information models usually simply accumulate the information of different evaluation indicators, without paying attention to the weight differences between each evaluation indicator, resulting in inaccurate susceptibility zoning results. This article takes the Zigui to Badong section of the Three Gorges Reservoir area as an example and proposes a landslide hazard susceptibility evaluation method based on random forest weighting information to improve evaluation accuracy. This method first uses factor feature analysis to determine landslide evaluation factors, and then uses a random forest model to determine the weights of each evaluation factor. Then, the weights are fused with an information model, and a more accurate susceptibility evaluation result is obtained by weighting and superimposing the information of the evaluation factors. Two models were analyzed and evaluated using statistical indicators and ROC curves. The results showed that the area under the ROC curve (AUC) of the traditional information model and the information model based on random forest weighting were 0.778 and 0.855, respectively, in the test set. The method of determining the weighted information of the random forest optimized the traditional information method, providing a new approach for landslide disaster risk assessment.