A prediction method for older adult care service demand combining improved RF algorithm and logistic regression

Area of Science:

• Gerontology
• Health Informatics
• Predictive Analytics

Background:

• Global aging populations necessitate precise demand forecasting for elder care services.
• Traditional prediction models struggle with complex, high-dimensional health data.

Purpose of the Study:

• To develop an advanced prediction model for older adults' care service demand.
• To improve resource allocation efficiency in aging societies.

Main Methods:

• An integrated model combining an improved random forest (RF) with logistic regression (LR).
• Incorporation of an adaptive feature selection strategy within the RF framework.
• Utilizing optimized RF-selected features to build an interpretable LR classifier.

Main Results:

• The integrated model achieved 95.30% accuracy, 92.60% recall, and an F1 score of 93.90%.
• Demonstrated superior performance compared to standalone RF or LR models.
• Achieved an Area Under the Curve (AUC) of 0.934.

Conclusions:

• The hybrid RF-LR model significantly enhances prediction accuracy and reliability for elder care demand.
• Provides a robust decision-support tool for optimizing care service planning and resource allocation.
• Addresses the challenges of high-dimensional, nonlinear health data in gerontological research.