Research On Landslide Hazard Zoning Based On Fuzzy C-means Clustering Method

Landslide disaster,as a destructive geological disaster,has been restricting the sustainable development of the mountains of China.After years of research and exploration,the study of landslide disaster has achieved fruitful results.However,comparing with the most advanced foreign research level,there is still a big gap.There are many problems in landslide hazard zoning,such as quantitative analysis,evaluation method and result test,and it is worth to study deeply.Many scholars,both at home and abroad,use unique condition units or slope units as mapping units for landslide hazard zonation.However,the unique conditional unit is easy to carry into the quantization error.and the slope unit is not enough to describe the horizontal surface,so the result is very different from the actual result.To this end,it is necessary to propose a reasonable method of applying the mapping unit.ROC analysis usually uses pairing design to verify the multi-classification model,so the model test calculation is more complex and less reliable.Therefore,it is necessary to improve the ROC analysis to verify the multi-classification validation and reduce the computational complexity of ROC analysis.The existing zoning studies one-sided pursuit of the complexity and advanced nature of the model,while ignoring the complexity of landslide hazards,is not be a good solution to the evaluation problem.On the basis of collecting relevant information about the detailed investigation of geological disasters in Jinyang County in 2015,this paper taking Jinyang County as an example,a new unit is presented,which combines the unique condition unit and the slope unit,and on the basis of the study of factor superposition division model,a fuzzy C-means clustering model is introduced,and based on the ROC curve analysis and the concept of membership degree in fuzzy mathematics theory,the ROC curve is applied to the accuracy test of multi-classification model in this paper.In conclusion,this paper establishes a model of landslide risk assessment based on fuzzy C-means clustering algorithm.Based on the combination of-unique condition unit and slope unit,eight influencing factors such as elevation and structure are selected as the evaluation factors,Fuzzy C-means clustering algorithm is used as the evaluation model to classify the landslide hazard zonation,and it is found that this method can reduce the unique conditional element quantization error and reduce the impact on the level of the ramp surface unit.The fitting degree and precision of ROC analysis can reach to 0.918 and 0.856,respectively.so it can prove the rationality of the method.By comparing the factor superposition method,we can prove the reliability of the ROC analysis based on the fuzzy mathematics theory in the multi-classification model.This paper involves three innovations as follows:1.The unit division method of the unique slope unit is established.The mapping units are divided into two parts,namely the calculation units and evaluation units.The calculation units are referred to the risk quantization units,and the evaluation units are referred to the risk division units.In this paper,unique condition units are used as the calculation unit,and slope units are used as the evaluation unit.Finally,both of them form the unique slope units.2.A method of landslide hazard zoning based on fuzzy C-means clustering method is introduced.The method is based on the similarity of the evaluation factors in the unit,and the correlation between the evaluation factors is low.3.An accuracy analysis method of ROC curve model with fuzzy membership degree is established.The fuzzy mathematics theory pushes the ROC analysis to high-dimensional(three-classification or more)applications,so that the ROC analysis for multi-classification model is simplified,and it will be used to analysis landslide hazard zonation models in this paper.