SOM: Algorithm of Spatial Outliers Mining Based on MST Clustering

摘要：

Spatial outliers are spatial objects whose non-spatial attributes deviate so much away from other datas in the datasets, spatial outliers mining can bring a lot of interesting information to us. However, because of some complicated characteristics of spatial data, such as topological relation, orientation relation, measurement relation,and so on. Traditional algorithms for outliers mining in business database are shown to unfit to spatial datasets. Based on clustering algorithm for partitioning a minimum spanning tree, a new algorithm of spatial outliers mining(SOM) was proposed. SOM algorithm keeps basic spatial architectural feature of spatial object by used of MST, and gets MST clustering by cutting off the most inconsistent edges, so the algorithm not only owns the characteristics that it can efficiently get clusters from non-spherical and uneven datasets like algorithms of density-based clustering, but also have the advantage that it dont depend on the parameters selected by user, so the result of clustering is usually more reasonable to us. In the end, the validity of SOM algorithm is validated through the real data.

关键词： SOM algorithm spatial outliers cluster-based outliers MST clustering

作者: Jia-Xiang Lin Chong-Cheng Chen Ming-Hui Fan Min-Qi Zheng

作者单位: Spatial Information Research Center of Fujian, Fuzhou University, Key Lab Of Spatial Data Mining & Information Sharing of Ministry of Education, Fujian, Fuzhou, 350002

会议类型: 国际会议

会议名称: 北京国际地理信息系统学术讨论会第七届会议（7th International Workshop Geographical Information System

会议地点: 北京

会议语种:英文

页码: 134-140

在线出版日期: 2007-09-14（万方平台首次上网日期，不代表论文的发表时间）