Policy Decision-Making for Population Decline Using AI to Estimate Population Density From Well-Being Indicators

Abstract

This paper examines whether a model that infers habitable area population density from regional well-being indicators can serve as a guide for policy decision-making to address population decline. The study uses 51 subjective evaluation items from the regional well-being indicators and habitable area population density calculated from e-stat, a Japanese government database. The inference model was created through ensemble learning, generating six weak learners and combining them with a meta-model to form the final model. Using data from Shimonoseki City in Yamaguchi Prefecture, Japan, we varied the value of a single subjective evaluation item to observe changes in the inferred population density. The results showed that subjective evaluations related to public transportation, crime prevention, dining options, and local government initiatives significantly impact habitable area population density. Prioritizing these factors could enhance resident satisfaction and potentially mitigate the issue of population decline.