Rsesearch And Application Of Uncertain DM-chameleon Clustering Algorithm In Landslide Hazard Prediction

Landslide is one of the most frequently occurring geological disasters in China even the whole world,it is not only destructive to property,environment and resources,but also poses a serious threat to human life safety.The research of landslide hazard risk prediction is a complex process in consideration of multi-source information,the information of that is of abruptness,non-linearity,randomness and uncertainty etc,which brings the corresponding difficulty to the landslide forecast.Due to the frequent occurrence of landslide and the more and more serious harm situation,the research topic of seeking a scientific and effective method to improve the accuracy of landslide prediction is becoming extremely meaningful.The theory of data mining has a strong ability to deal with nonlinear relationships.For example,clustering methods can be used to classify highly similar data objects in the same class,and the highly different ones are divided into different classes then unknown things will be predicted according to the known rules.So,in this paper we use the Chameleon clustering algorithm and combine the characteristics of landslide information to build the landslide hazard prediction model and propose the clustering subclass risk classification method,both of which are used to predict the risk classification of landslide in the study area.The study found that landslide caused by a variety of factors,such as slope height,slope type,rainfall and so on,the rainfall is an important factor for the landslide hazard prediction.It is an uncertain attribute whose value is between a range not an exact value,which is difficult to be effectively processed by the traditional Chameleon clustering algorithm.And the traditional Chameleon clustering algorithm also exists some problem.For example,some parameters should be selected by hand when using it to choose k value and the threshold value of similarity degree function,and it is unable to deal with large-scale data sets.In order to solve the above problems,based on M-chameleon clustering algorithm of the previous research,this paper proposed an DM-chameleon clustering algorithm and introduced an uncertain data models to describe uncertain factors such as rainfall.The new algorithm is able to deal with uncertain data by extending the Euclidean distance of similarity in cluster analysis.Finally,obtained uncertain DM-chameleon clustering algorithm and its application in the landslide risk prediction model,selected Baota district of Yan'an study area to verify this method.Firstly,experiments on the known data sets show that DM-chameleon algorithmobtains better clustering effect than M-chameleon algorithm and the clustering speed has been improved obviously.Then compare the experimental results show that the uncertain DM-chameleon clustering model achieved a higher prediction accuracy,verified the feasibility of uncertainty DM-chameleon clustering algorithm in landslide hazard prediction.