Ensemble Learning for Disease Prediction: A Review.

Area of Science:

• * Computational biology and bioinformatics
• * Medical informatics and machine learning
• * Health data analytics

Background:

• * Machine learning models enhance disease prediction frameworks.
• * Ensemble learning combines multiple classifiers for improved accuracy.
• * Limited assessment of ensemble methods for common diseases exists.

Purpose of the Study:

• * To assess ensemble techniques (bagging, boosting, stacking, voting) for disease prediction.
• * To identify performance trends against five major diseases: diabetes, skin, kidney, liver, and heart conditions.
• * To provide insights for selecting optimal predictive models.

Main Methods:

• * Systematic literature search for studies from 2016-2023.
• * Identified 45 articles applying at least two ensemble methods to the target diseases.
• * Comparative analysis of performance accuracies across different ensemble approaches.

Main Results:

• * Stacking, despite fewer applications (23), yielded the highest accuracy most often (19/23).
• * Voting was the second-best performing ensemble method.
• * Bagging excelled in kidney disease prediction; boosting in liver and diabetes.
• * Stacking consistently showed top performance for skin and diabetes.

Conclusions:

• * Stacking demonstrates superior accuracy in disease prediction over other ensemble methods.
• * Performance varies significantly across different ensemble approaches and disease datasets.
• * Findings guide researchers in selecting appropriate ensemble models for predictive analytics.