Classification and Space Cluster for Visualizing GeoInformation

Abstract

It is necessary to classify numerical values of spatial data when representing them on a map so that, visually, it can be as clearly understood as possible. Inevitably some loss of information from the original data occurs in the process of this classification. A gate loss of information might lead to a misunderstanding of the nature of original data. At the same time, when we understand the spatial distribution of attribute values, forming spatial clusters is regarded as an effective means, in which values can be regarded as statistically equivalent and distribute continuous in the same patches. In this study, a classification method for organizing spatial data is proposed, in which any loss of information is minimized. Also, a spatial clustering method based on Akaike's Information Criterion is proposed. Some numerical examples of their applications are shown using actual spatial data for the Tokyo metropolitan area.