A Predictive Model for Guillain-Barré Syndrome Based on Ensemble Methods.

Area of Science:

• Computational neuroscience
• Medical informatics
• Machine learning applications in healthcare

Background:

• Machine learning (ML) is effective for disease prediction, yet Guillain-Barré syndrome (GBS) lacks computational study.
• Previous work utilized single classifiers for GBS prediction; a robust model is needed for timely patient treatment.

Purpose of the Study:

• To develop and evaluate a novel predictive model for classifying four subtypes of Guillain-Barré syndrome.
• To compare the performance of ensemble methods against single classifiers for GBS subtype identification.

Main Methods:

• Three classification experiments were conducted: all GBS subtypes, One versus All (OVA), and One versus One (OVO).
• A real-world dataset of 129 instances with 16 features was used.
• Five ensemble methods were compared against 15 single classifiers over 30 independent runs.

Main Results:

• Random Forest demonstrated superior performance in classifying the four GBS subtypes.
• No single ensemble method outperformed others in the OVA classification experiment.
• Single classifiers generally outperformed ensemble methods in the OVO classification scenario.

Conclusions:

• A novel predictive model for Guillain-Barré syndrome subtype classification has been presented.
• The model identifies optimal classification methods for different scenarios, aiding physician decision-making.
• This work provides a foundation for future, improved predictive models for GBS.