Long-Range Prediction of Population by Sex and Age in Each District
Based on Fuzzy Theory

Abstract

This paper proposes a method of predicting the population by sex and age for each of 402 districts over a long-range period in the Kansai region, Japan, by applying fuzzy theories. First, to predict the total social increase for 402 districts by directly taking into consideration of differences in factors of migration in each district, nine rules or domains were set up by using the migration rate and the total social increase in each district as the premise variables. Regression models were constructed in the consequences which use various socioeconomic indicators as explaining variables. The future value of the total social increase in each district can be obtained by weighting the values calculated from the estimated regression models with the membership values denoting the degree of belonging to each rule. Second, a method to estimate the social increase by sex and age in each district is proposed based on fuzzy clustering method for dealing with long-range socioeconomic changes in population migration by sex, age and district. All the samples of the migration ratio were classified into the same nine domains. By applying Fuzzy c-Means on districts belonged to each domain, all samples were classified into 20 clusters. The future migration ratio in each district can be estimated by weighting the migration pattern in each cluster with the values of membership function denoting the degree of belonging to each cluster. Results of the validity test of the constructed population model based on the proposed methods are also presented. It has been shown that it becomes possible to predict the population by sex, age and district over a long-range period by using the proposed method.