Abstract
We propose a method to visualize features of geographical regions using origin-destination matrices. The spatial proximity such as the distance is a major factor affecting people's locational decision making. This information is reflected in the OD matrix, and thus, analyzing relationships between regions without distance data is an interesting topic to explore. Based on the similarity between rows of the OD matrix, we formulate optimization models to visualize regions in a two-dimensional space where regions with similar characteristics are placed close to each other, and regions with low similarity are placed far from each other. We apply the method to the spatial flow data of the admission to seven national universities from 47 prefectures in Japan. The result illustrates similarities to the actual position of Japanese prefectures and reveals many geographical findings which are difficult to obtain from the raw data. We also propose a method of dividing prefectures into a specified number of groups by solving the median problem.
