Abstract
Ordinarily, when we divide a whole region into some homegenious subregions, cluster analysis is often used. But, by the application of the ordinary cluster analysis, such a division that unit districts belonging to a cluster form a subregion which is composed of only one locational block cannot be realized. According to the object of the division, there are divisions which have to satisfy such a condition. Then, in this paper, the author tried to devise such a dividing method that satisfies the condition by the application of the cluster analysis, and applied the method to the division of the all 98 municipalities in San-in region from a standpoint of agricultural homegeneity.
In order to satisfy the condition, it become necessary to judge whether any two clusters are neighboring to each other or not. Then, in this paper, the judegement was accomplished based on whether any two unit districts which belong to different clusters to each other are neighboring or not. Only the pairs of two clusters whose unit districts satisfy this criterion become subjects of the fusion at each fusing stage of the clusters. Other fusing processes are the same with those of the ordinary cluster analysis. By the application of the method, the author divided the above 98 municipalities into 15 groups (clusters) which satisfy the condition.
The author also presented the method in the case that there are differences in the degree of the contiguities between any two unit districts. In this case, two ranks of the degree were set up.
The method which is devised in this paper is important and effective if unit districts belonging to a cluster have to form a subregion which is composed of only one locational block. Moreover, the method has a capability that we can add other various conditions for the fusion. But, the establishment of the conditions means that strict conditions are imposed on the fusing process. Then, we have to be careful that more subregions are necessary if we need similarity among unit districts belonging to a subregion as much as that in the case of the ordinary cluster analysis.
