Data underlying the publication: Optimizing multi-resolution topographic indicator combinations for debris flow susceptibility assessment based on factorial experiments and UMAP analysis

This dataset supports the study "Optimizing multi-resolution topographic indicator combinations for debris flow susceptibility assessment based on factorial experiments and UMAP analysis". It leveraging a factorial experimental design and UMAP analysis to systematically evaluated nearly 350,000 prediction results from RF, GBDT, and BPNN models. Then a multi-faceted assessment across seven dimensions revealed the impact of resolution combinations on susceptibility and identified each model's optimal combination. The dataset contains the basic training and testing sets used in this study to assess debris flow susceptibility.

本数据集支撑了题为“基于因子试验与UMAP分析的泥石流敏感性评估多分辨率地形指标组合优化”的研究。本数据集采用因子试验设计与UMAP分析方法，系统评估了随机森林（Random Forest，RF）、梯度提升决策树（Gradient Boosting Decision Tree，GBDT）以及反向传播神经网络（Back Propagation Neural Network，BPNN）三类模型的近35万条预测结果。随后通过覆盖七个维度的多维度评估，揭示了分辨率组合对泥石流敏感性评估的影响，并确定了各模型的最优分辨率组合。本数据集包含本研究中用于泥石流敏感性评估的基础训练集与测试集。