论文部分内容阅读
以四川省低山丘陵区为研究区,基于滑坡编目数据,在深入分析研究区滑坡孕灾环境的基础上,选取坡度、地形起伏度、岩土类型和断裂构造等八类孕灾环境因子。通过融合确定性系数和多层感知器,提出CF-MLP模型,并对研究区滑坡的敏感性进行评价,计算得到的滑坡敏感性指数在ROC曲线中的线下面积为0.84,说明该模型预测结果对滑坡具有较好的识别作用;基于模型预测结果对研究区进行敏感性区划,共分为高敏感性、中敏感性和低敏感性三个区域,与历史滑坡的分布现状相一致。实践证明,CF-MLP模型在一定程度上解决了滑坡孕灾环境因子数据的合理量化问题,提高了多层感知器网络的收敛效果,有利于建立更为准确的滑坡敏感性分析模型。
Based on the cataloging data of landslide in Sichuan Province, based on the cataloging data of landslide, eight types of environmental factors of pregnancy disaster including slope, topographic relief, rock type and fault structure were selected based on the analysis of the landslide pregnant environment in the study area. The CF-MLP model is proposed by integrating deterministic coefficient and multi-layer perceptrons. The sensitivity of the landslide in the study area is evaluated. The calculated area under the line of ROC curve of the landslide sensitivity index is 0.84, indicating that the model predicts The result is good for landslide identification. Based on the model predictions, the sensitivity zoning of the study area is divided into three areas: high sensitivity, medium sensitivity and low sensitivity, consistent with the distribution of historical landslides. The practice proves that the CF-MLP model solves the problem of quantitatively quantifying environmental factors of landslide hazard to some extent, improves the convergence effect of multi-layer perceptron network, and helps to establish a more accurate slope sensitivity analysis model.