Research And Application Of Uncertain UGM-CFSFDP Algorithm In Landslide Hazard Prediction

Landslide pose a great threat to the environment and resources,and hava a great impact on people's lives.The occurrence of landslides is affected by various precipitating factors like rainfall.The rainfall is an uncertain facter,and the traditional clustering methods cannot accurately process the rainfall,which brings corresponding challenges to landslide hazard prediction.Therefore,it has a very important practical significance to find a scientific and effective forecasting method.The clustering algorithms in data mining can classify data with higher similarity into one class and predict unknowns based on known information.In this article we use clustering by fast search and find of density peaks(CFSFDP algorithm)as the basic algorithm.Introduce the t-nearest neighbor information to define the relative density as a new measure of the object point density.optimize the selection of cluster centers in the data set,overcome the problem that the selection of the cluster center depends on the defect of setting density thresholds when data distribution is not uniform.Introduce the idea of meshing and merging to avoids the limitation of using global thresholds in clustering and realize effective processing of large-scale data.According to the above theory,we propose a GM-CFSFDP clustering algorithm which can realize the processing of largescale and uneven density data sets.Considering the characteristics of landslide hazards comprehensively,we build a landslide hazard prediction model to predict landslide hazard rating.Landslide caused by a variety of factors,the rainfall is an important factor whose value is between a range not an exact value,which is difficult to be effectively processed by the GMCFSFDP clustering algorithm in landslide risk prediction.In order to solve the problem that rainfall and other uncertainties are difficult to handle effectively,Firstly introduce the uncertain data model,consider the difference between points in the interval range,extend the traditional Euclidean distance,the E-ML distance is designed which can effectively handle the uncertain data such as rainfall.Then,we propose the uncertain UGM-CFSFDP algorithm by introducing the EML distance into the GM-CFSFDP algorithm.The experiments results on uncertain simulation data sets show that the uncertain UGM-CFSFDP algorithm has higher clustering quality.Finally,we build the landslide risk prediction model of uncertain UGM-CFSFDP algorithm to predict Baota district of Yan'an study area.The result proves that the uncertain UGM-CFSFDP algorithm can obtain better prediction accuracy,thus verifying the feasibility and advancement of the algorithm in landslide hazard prediction.