Optimizing brain stroke detection with a weighted voting ensemble machine learning model

Area of Science:

• Neurology
• Computer Science
• Medical Informatics

Background:

• Brain stroke is a critical medical condition caused by disrupted blood flow, leading to brain cell death.
• Current diagnostic methods like CT scans and MRIs are often time-consuming and expensive.
• Early stroke risk diagnosis is crucial for timely preventive actions.

Purpose of the Study:

• To develop an ensemble machine learning model for accurate and efficient brain stroke prediction.
• To improve upon existing stroke diagnostic methods by leveraging artificial intelligence.
• To enable early identification of stroke risk for proactive healthcare.

Main Methods:

• Developed a weighted voting-based ensemble (WVE) classifier.
• Integrated multiple individual classifiers: random forest, eXtreme gradient boosting, and histogram-based gradient boosting.
• Trained and evaluated the model on a private stroke prediction dataset.

Main Results:

• The proposed WVE ensemble model achieved a high accuracy of 92.31% in predicting brain stroke.
• The ensemble approach demonstrated superior performance compared to individual classifiers.
• The model offers a feasible solution for early stroke diagnosis.

Conclusions:

• Ensemble machine learning models can effectively predict brain stroke risk.
• The developed WVE model provides a promising, cost-effective, and timely alternative to traditional stroke diagnostics.
• Further research into intelligence-based optimization can enhance model accuracy and clinical utility.