摘要
针对滑坡危险性预测中降雨等不确定因素难以衡量,及现有的预测方法大多属于无监督的传统聚类方法,不能有效利用先验信息的问题,为有效提高预测精度,首先提出一种不确定数据距离-uv距离,它实现了不确定因素降雨的有效刻画;其次将半监督聚类应用于滑坡危险性预测,引入uv距离,设计了一种基于不确定数据的半监督动态K-均值算法,其有效利用了先验信息,并通过设置隶属度阈值实现了数据集的动态划分,有效提高了预测精度。研究区的实验结果证明了uv距离及算法的有效性。
Due to the difficulties of the proper measurement for the uncertain factor such as rainfall and the effective use of the prior knowledge for the traditional unsupervised clustering method in the landslide hazard prediction, the accurate prediction of landslide hazard has not been achieved. In this paper, a uncertain data distance--uv distance which effectively measures the uncertain factor is proposed, then through the creative idea that introducing the semi-supervised method in landslide hazard prediction, a uncertain semi-supervised dynamic algorithm based on the uv distance is perfectly designed,which fully uses the prior information and achieves the dynamic division of the data set by the membership threshold.Experimental results in the study area demonstrate the effectiveness of the uv distance and the algorithm.
作者
朱玲
ZHU Ling(Jingdezhen University,Jingdezhen 333000,China)
出处
《山东农业大学学报（自然科学版）》
北大核心
2020年第2期340-346,共7页
Journal of Shandong Agricultural University：Natural Science Edition
关键词
不确定数据
半监督聚类
危险性预测
滑坡
uncertain data
semi-supervised clustering
hazard prediction
landslide
