Abstract
In this paper, the spatial distribution of business composition is estimated from Yellow pages database of 1990, 1995 and 2000 in Tokyo 23 wards area. First, using MDL criterion and binary tree segmentation, we created suitable area-partitioning model for position data and attribute data of business type. And business composition of each area segment is estimated, and Kullback-Leibler information, which acts like a distance measure between two distributions, shows how it is different from average composition in all 23 wards. Then spatial distribution of each business type is also estimated. Characteristic districts such as Akihabara are found on the map colored by the information-measure.
