Prediction of successful aging using ensemble machine learning algorithms.

Area of Science:

• Gerontology
• Biomedical Science
• Computational Science

Background:

• Aging is a primary risk factor for chronic diseases, increasing healthcare costs and societal burdens.
• Successful aging (SA) is defined biomedically as the absence of disease and disability, contrasting with normal aging's functional decline.
• Socially, SA emphasizes life satisfaction and well-being, though its definition and measurement remain imprecise.

Purpose of the Study:

• To identify the most effective features of successful aging based on Rowe and Kahn's theory.
• To develop and validate machine learning (ML) predictive models for successful aging using identified features.

Main Methods:

• A retrospective study involving 983 participants.
• Training of six ML algorithms: artificial neural network, decision tree, support vector machine, Naïve Bayes, and k-nearest neighbors (K-NN), including an ensemble of 30 K-NN algorithms.
• Prediction using a majority vote method based on the ensemble model's output.

Main Results:

• The predictive system achieved high performance metrics: 93% precision, 92.40% specificity, 87.80% sensitivity, 90.31% F-measure, 89.62% accuracy, and 96.10% ROC.
• A five-fold cross-validation procedure was employed for model evaluation.
• The K-NN-based ensemble algorithm demonstrated superiority in classifying individuals into successful aging and non-successful aging categories.

Conclusions:

• Machine learning techniques show significant potential in supporting policy decisions related to successful aging.
• The K-NN-based ensemble model is highly effective for classifying individuals based on successful aging criteria.
• These findings can inform health and social policymakers for better resource allocation and interventions for aging populations.