Construction of a Regional Healthcare Model Using Bayesian Networks and Open Data

Abstract

In Japan, it's required to utilize local government health data, but challenges arise due to its vast scale and the complexity of creating understandable models. A study employed Bayesian networks for regional medical analysis, using U.S. open data for evaluation. While networks built with feature selection algorithms were less accurate than fully connected ones, they were better in terms of model complexity. Interestingly, the model revealed a higher proportion of high blood pressure among non-smokers, possibly due to nicotine's effects on appetite and stress. Future analysis will use KDB to understand regional health patterns.