A Study on an AI-Based Model for Predicting Elderly Self-Reported Health Status: Analysis Using Korea’s Nationally Representative Triennial Elderly Survey
Keywords: Elderly Health Assessment, AI-Driven Health Prediction, National Survey of the Elderly (Korea), Random Forest, XGBoost, Synthetic Minority Over-sampling Technique (SMOTE
Submission Type: Abstract
Status: In Review | Submitted at: 2025-05-23 23:39:53
Abstract
This study aims to develop and validate an AI-based model for predicting elderly individuals' self-reported health status using data from the 2023 edition of the triennial National Survey of the Elderly conducted by the Ministry of Health and Welfare of the Republic of Korea. Unlike previous studies that primarily utilized hospital records or wearable sensor data, this study constructs an AI model by integrating large-scale survey data with clinically diagnosed health assessment data obtained through comprehensive medical check-ups. This approach significantly enhances predictive accuracy and real-world applicability. A predictive model was developed by extracting 135 health-related variables from the survey dataset and classifying respondents' self-reported health status into six categories. The AI models were trained using Random Forest and XGBoost, while data imbalance issues were addressed through the implementation of the Synthetic Minority Over-sampling Technique (SMOTE). Performance analysis demonstrated that both models achieved high classification accuracy, with AUC values of 0.89 or higher. These results indicate that the proposed model provides reliable predictions of elderly health status, supporting its practical applicability. The proposed AI-based model offers a novel approach to analyzing elderly individuals' self reported health status, contributing to the advancement of personalized healthcare management and improved medical service delivery. Furthermore, integrating this model with digital healthcare systems enhances its usability for early intervention strategies and provides valuable insights for policymakers in designing more effective elderly welfare policies.
Authors
- Kyung-Hyun Lee (First Author), Department of IT·Semiconductor Convergence Engineering, Sangidaehak-ro, Siheung si, Gyeonggi-do, Korea, Republic of Korea 2Catholic Kwandong University Industry Cooperation Foundation – khlee@cku.ac.kr
- Rackjune Baek (Co-author), CKU Professor – rj100@hanmail.net