Stroke Disease Detection and Prediction Using Robust Learning Approaches.

Area of Science:

• Neurology
• Medical Informatics
• Data Science

Background:

• Stroke is a leading global cause of death and disability.
• Prompt identification of stroke warning signs is crucial for mitigating severity.
• Machine learning (ML) offers potential for predicting stroke likelihood.

Purpose of the Study:

• To develop and compare ML models for reliable stroke prediction.
• To evaluate the performance of Logistic Regression, Decision Tree, Random Forest, and Voting Classifier algorithms.
• To identify the most accurate ML algorithm for stroke risk assessment.

Main Methods:

• Utilized physiological parameters and an open-access Stroke Prediction dataset.
• Trained four distinct ML models: Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), and Voting Classifier.
• Compared model performance based on prediction accuracy.

Main Results:

• Random Forest (RF) classification demonstrated the highest performance.
• RF achieved an accuracy of approximately 96 percent in stroke prediction.
• The developed models significantly outperformed previous studies in accuracy.

Conclusions:

• Machine learning, particularly Random Forest, provides a robust and reliable method for stroke prediction.
• The high accuracy of these models can aid in early stroke detection and intervention.
• Further research can build upon these findings for enhanced stroke prevention strategies.