The research focused on predicting strokes, a significant threat to health and well-being. The primary challenge addressed was the use of a highly imbalanced dataset. Various data preprocessing techniques were employed to tackle this, enabling the construction and comparison of machine-learning models for stroke prediction. Among the models assessed, the random forest model proved to be the most effective, achieving precision, recall, and F1-score levels of 90%, along with an accuracy of 90%. A random forest classifier was also trained using optimal hyperparameters obtained via grid search on balanced data to highlight the limitations of relying solely on accuracy in classification tasks. This approach demonstrated the model's high accuracy of 96%, underscoring the impracticality of using accuracy as the sole metric for performance evaluation in imbalanced datasets. In conclusion, the research underscores the critical role of advanced data processing and machine learning techniques in enhancing stroke prediction. The successful application of the random forest model and the recognition of accuracy's limitations provide a robust framework for future studies and practical implementations in healthcare settings. This work advances the predictive capabilities for stroke and paves the way for improved approaches to healthcare strategies, ultimately aiming to save lives and enhance the quality of life for individuals at risk.
Melnykova et al. (Tue,) studied this question.