Optimization Of Uncertain Density Clustering Algorithm For Landslide Data

Landslides are a type of geological disaster that is extremely destructive and frequent,and will cause extremely serious economic losses and casualties every year in our country.Therefore,it is necessary to predict the risk of landslides in combination with landslide hazard data and thus to carry out targeted landslide prevention and control work.Density clustering algorithm is a key technology in cluster analysis which has the ability of taking account of many key factors that affect the development of landslides,and usefully avoid the influence of cluster shape on the prediction results of landslides,so as to effectively classify landslides data and extract potential useful information.However,it is difficult to effectively describe the rainfall,requires the user to input the density threshold,and has high time complexity when the traditional density clustering algorithm is applied to landslide hazard prediction.In view of the shortcomings mentioned above,this paper improves on the basis of OPTICS-PLUS clustering algorithm and ASCABHD clustering algorithm proposed by predecessors.The main work is as follows:1)For traditional density clustering algorithms,it is difficult to effectively deal with the uncertain data in landslide hazard data.Considering the distribution characteristics of uncertain data in the range,the cloud model theory is introduced into the EW distance formula,then,an EC type distance formula is proposed to effectively handle uncertain data such as rainfall.2)Aiming at the deficiencies of OPTICS-PLUS algorithm in landslide disaster prediction,the concepts of point average distance and nearest neighbor distance are put forward based on OPTICSPLUS algorithm to optimize the expansion strategy and storage structure,and then,a NNSBOPTICS clustering algorithm is designed,which overcomes the insufficiency of the time efficiency and the shortcoming of requires the user to set a density threshold.EC-distance formula is used as the similarity measure formula,which is applied to NNSB-OPTICS algorithm,and the uncertain NNSB-OPTICS clustering algorithm is proposed,which is suitable for clustering datasets containing uncertain data.3)Aiming at the deficiencies of ASCABHD algorithm in landslide hazard prediction application,a KDEB-PDCA clustering algorithm is designed.This algorithm combines the kernel density estimation method to obtain the peak density object,avoids the user from setting the density threshold,divides the data set into several grids,and adopts the idea of divide and conquer for clustering,which effectively improves the time efficiency.In order to effectively deal with uncertain data in landslide hazard prediction,the EC type distance formula is introduced into the KDEB-PDCA algorithm,and an uncentain KDEB-PDCA clustering algorithm is proposed.The simulation experiment on UCI dataset verifies that the proposed NNSB-OPTICS algorithm and KDEB-PDCA algorithm have high clustering accuracy and stability,and they have certain advantages in time efficiency.Based on the field survey data of landslide disaster in Baota District of Yan'an,combined with the uncertain NNSB-OPTICS algorithm and the uncertain KDEB-PDCA algorithm proposed in this paper,a landslide hazard prediction model is established and a landslide hazard prediction experiment is conducted to prove that EC distance formula proposed in this paper can effectively process the uncertain data of rainfall in landslide data,thereby improving landslide prediction accuracy.