Federated Learning-Based AI Approaches for Predicting Stroke
Year: 2024
Type: book-chapter
Abstract: In recent years, rapid growth and extensive research have driven the proliferation of machine learning (ML) applications. This is particularly vital in addressing the global public health crisis of stroke, the second leading cause of death worldwide according to the World Health Organization. This study employs ML to create a precise model for predicting stroke outcomes using past data and individual characteristics, even on an imbalanced dataset. Utilizing a dataset from Kaggle containing diverse patient medical variables, four distinct algorithms are implemented and compared to enhance accuracy. The aim is to predict stroke occurrences to enable timely intervention and prevent devastating consequences. Addressing the dataset’s class imbalance, where ‘regular’ instances dominate, preprocessing is performed to handle missing data. Three classifiers, Extreme Gradient Boost (XGBoost), Random Forest (RF), and LightGBM, are employed individually, and an ensemble method combining these classifiers is used. Evaluation metrics such as accuracy, F1-score, recall, and precision are employed for performance measurement. This research identifies critical parameters for stroke prediction, with age, heart disease, average glucose level, and hypertension emerging as the most significant indicators. Comparing the four models, RF yields the best results, surpassing existing methods in terms of performance.
Source: CRC Press eBooks
Language: English
Subfield: Health Information Management
Field: Health Professions
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W4401097652