Health Prediction for Rural Older Adults Based on Artificial Intelligence Deep Neural Network

With the intensification of global population aging, health management for the elderly in rural areas faces challenges such as insufficient medical resources, outdated service facilities, and inadequate professional capacity. Against this backdrop, the early warning of health risks primarily related to hypertension and cardiovascular chronic diseases, as well as the real-time assessment of health status based on multi-source physiological data, have become urgent issues to address. To address this challenge, this study proposes an Artificial Intelligence-Driven Deep Neural Network (AI-DNN) health prediction model designed for rural intelligent elderly care scenarios. This model fuses multi-source heterogeneous physiological and behavioral data. It adopts an improved residual gated recurrent unit combined with a two-level attention mechanism for health risk perception to jointly model key time segments and high-risk physiological indicators. Meanwhile, an adaptive feature reweighting module is designed to dynamically adjust feature contributions according to individual health status. This enhances the model's representation capability and predictive performance for sparse, noisy, and heterogeneous health data. On the large-scale real-world rural health monitoring dataset covering 2158 elderly residents in several administrative villages in three provinces, the AI-DNN model achieves 92.3% accuracy, 91.7% F1 value and 0.948 AUC, which was significantly better than the baseline model, and p < 0.05. In addition, the parameter of AI-DNN model is 9.3M, which is in the middle and low level among all depth models, and it is a lightweight structure with low reasoning delay. This study provides deployable and robust technical support for intelligent elderly care systems in rural China, and provides useful solutions and experience for elderly health management in other resource-limited regions around the world. The study has broad application potential and promotion value.