Leveraging Adaboost and Gridsearchcv for Optimized Stroke Prediction Outcomes
Loading...
Date
2025-04-08
Journal Title
Journal ISSN
Volume Title
Publisher
Sule Lamido University Journal of Science and Technology
Abstract
Background: Globally, stroke is one of the most common causes of mortality and long-term disability. According
to World Health Organization (WHO) report, 15 million people suffer a stroke annually, with 5 million dying and
another 5 million becoming permanently disabled, placing a major burden on families and communities. Some
researchers have attempted to predict stroke using machine learning algorithms; however, while these algorithms
yielded high accuracy, limited attention was given to hyperparameter tuning and class imbalance issues. Aim:
This study addresses the critical need for enhanced stroke prediction by employing the AdaBoost machine
learning algorithm, optimized through GridSearchCV for hyperparameter tuning to mitigate the aforementioned
problems. Method: Utilizing a dataset from Kaggle, pre-processing steps, including feature selection, encoding,
and class balancing via synthetic minority oversampling technique (SMOTE), were implemented. The model's
performance significantly improved post-tuning, with accuracy rising from 75.42% to 84.33%, precision from
71% to 79%, and recall from 87% to 93%. These results highlight the importance of hyperparameter optimization
in enhancing machine learning models for medical diagnostics. Results: The findings suggest that the optimized
AdaBoost model can aid in early stroke detection, potentially improving patient outcomes and offering a valuable
tool for healthcare providers. This study contributes to the field by setting a benchmark for stroke prediction and
providing a foundation for future advancements in predictive analytics.
Keywords: prediction, adaboost, gridsearchcv, machine learning, hyperparameter tuning