期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2002
卷号：XXXIV Part 2
页码：443-448
出版社：Copernicus Publications
摘要：Some new methods for analyzing geo-referenced statistical data are presented in this paper. These methods have combined the techniques of exploratory data analysis with algorithms for data mining. They have been integrated in a prototype software system developed at the Technical University of Munich in cooperation with Navigation Technologies (NavTech) GmbH. The system serves the purpose of value-adding the road database maintained by NavTech. In the original database, each road element is described by more than 110 attributes. A number of algorithms on the basis of information gain and rough-set modeling have been implemented to rank the individual attributes and detect the dependencies among attributes based on their values in an arbitrarily selected region. Other algorithms are developed on the basis of road geometry and devoted to the quantitative description of spatial patterns such as routes and urban structures. With the knowledge of relative importance of the individual attributes, users are given the flexibility to buy a local road database with truncated attribute list. By observing the ranking list and correlation matrix calculated for different regions, information that reflects the regional differences of a road network can be extracted. Likewise, the changes in ranking list and correlation matrix of the same region after removing or adding a route imply the relative importance of this particular route.
关键词：Data mining; Database; Navigation system; Information gain; Rough set modelling; Attribute dependence
