Integrated bagging-RF learning model for diabetes diagnosis in middle-aged and elderly population

Area of Science:

• Medical Informatics
• Machine Learning
• Public Health

Background:

• The aging global population presents challenges for healthcare resource allocation, particularly with the rising prevalence of diabetes.
• Accurate diabetes prediction models are essential for efficient healthcare resource management and timely patient intervention.

Purpose of the Study:

• To develop and evaluate an improved diabetes prediction model, the Bagging-RF model, specifically for middle-aged and older adults.
• To enhance the identification of diabetes in distinct age cohorts within the older adult population.

Main Methods:

• Preprocessing of two Kaggle diabetes datasets, including data cleaning, outlier removal, and age-specific screening (50-60, 60-70, 70-80).
• Application of the Synthetic Minority Over-sampling Technique (SMOTE) for data balancing.
• Training and comparison of the Bagging-RF model against eight other machine learning classifiers.

Main Results:

• The Bagging-RF model demonstrated superior performance, achieving high accuracy and F1 scores across all tested age groups (e.g., 97.35% accuracy and 97.35% F1 score for the 50-60 age group).
• The model consistently outperformed other integrated learning models, including ET, RF, Adaboost, and XGB, in diabetes prediction.
• The proposed model shows significant potential for accurate diabetes identification in the target demographic.

Conclusions:

• The Bagging-RF model is a highly effective tool for predicting diabetes in middle-aged and older adults.
• This predictive capability can significantly aid in optimizing healthcare resource allocation and improving patient outcomes.
• The study highlights the importance of tailored machine learning approaches for public health challenges in aging populations.