Predicting the Health Behavior of Older Adults in Western Hunan Villages Using Machine Learning Algorithms

Abstract

We extracted the spatial characteristics of the health-related behaviors of 1223 older adults of Tujia, Yao, Miao and Dong ethnicity living in 20 villages in western Hunan, considering three dimensions: spontaneously healthy, health-necessary, and mentally healthy behavior. We constructed separate prediction models using logistic regression, support vector machine, categorical boosting, random decision forest, light gradient boosting machine, and extreme gradient boosting. We then combined these models with Shapley additive explanations to complete a global explanatory analysis to explore the correlation between location and the health behaviors of older adults of different ethnicities living in villages. The support vector model and gradient boosting tree models produced the most accurate simulations of the health behaviors of older adults. We found significant differences in the health behaviors of the older adults in the different villages, noting the preferences of the older adults of specific ethnicities. This study provides a reference for the excavation of the health behavior of older adults and the aging design of village spaces.