The Semi-supervised Clustering Algorithm Based On The Uncertain Data For Landslide Hazard Prediction

China is a country with frequent landslide hazards.Every landslide will cause extremely serious damage and loss to our country's natural resources,human life and property.Therefore,it is important to seek landslide hazard prediction methods that can mitigate or even avoid disaster losses.Practical significance and value.The uncertainty of the main inducing factors of landslides often brings some difficulties to the landslide prediction.How to effectively deal with the uncertain data and design an effective landslide hazard prediction method is the focus of this paper.The clustering algorithm in data mining has been widely used in the field of landslide hazard prediction.The clustering algorithm belongs to the unsupervised algorithm.It clusters the data by the similarity between the data without any prior information.The traditional clustering method In the process of landslide hazard prediction application,a small amount of prior information is completely ignored.The clustering result is given significance only by using prior information after clustering.However,this landslide hazard prediction method usually has low prediction accuracy.And it is very likely that there will be classes that have no meaning.Therefore,considering the practical situation that a small amount of prior information is usually easily obtained in the practical application of landslide hazard prediction,in order to make full use of the given prior information,this paper first proposes the application of semi-supervised clustering method to landslide hazard prediction.Imagine,based on the more mature semi-supervised k-means algorithm based on seeds set,firstly realize the seeds set optimization based on the distribution characteristics of the landslide,and then use the seeds set as the initial class guide data set to perform similarity clustering.Further,the performance of the algorithm is effectively improved by setting the membership threshold.However,in the application of landslide hazard prediction,the rainfall of one of the main factors of landslide is uncertain data,and its value is usually within an interval.However,traditional methods are difficult to accurately measure the similarity between uncertain factors.Therefore,It can effectively describe the similarity of uncertain rainfall and then effectively cluster.This paper first proposes a concept of uncertainty data similarity,and draws on the idea of uniform distribution of uncertain data distance--Hausdorff distance.A wider uncertainty distance-uv distance is used as a metric to improve the similarity of uncertain data in semi-supervised K-means algorithm based on seeds set.The algorithm is applied to landslide hazard prediction and a kind of design is designed.Based on the semi-supervised k-means algorithm of uncertain data,the landslide hazard prediction model is used.Finally,the Yan'an Baota area is used as the research area,and the landslide hazard prediction model is applied to the pagoda area.The comparative test is used to verify the landslide hazard prediction.Supervised clustering is better than unsupervised clustering,and uncertain uv distance is used to measure uncertain rainfall.The validity and effectiveness of the semi-supervised k-means algorithm based on uncertain data on landslide hazard prediction are also verified.