Application of Synthetic Minority Over-Sampling Technique to Mitigate Class Imbalance in Stroke Prediction

Class imbalance in healthcare datasets poses significant challenges for predictive modelling, particularly in identifying minority class outcomes such as strokes. Machine learning models trained on imbalanced datasets, where one class significantly outnumbers others, often exhibit bias towards the majority class. This can severely impair the accurate identification of minority class instances. This study investigates the effectiveness of the Synthetic Minority Over-sampling Technique (SMOTE) in mitigating this bias, specifically within the context of stroke prediction models. To address the class imbalance within the dataset, SMOTE was applied to oversample the minority class. Subsequently, three machine learning models—Random Forest, Logistic Regression, and Support Vector Machines (SVM)—were trained and evaluated using a suite of performance metrics, including accuracy, precision, recall, F1-score, and the area under the Receiver Operating Characteristic (ROC) curve (AUC). The results indicate that the application of SMOTE significantly improved recall and F1-score for the minority class, effectively mitigating the challenges posed by the imbalanced class distribution. Random Forest consistently outperformed other models, while SMOTE yielded transformative improvements in Logistic Regression and SVM. This paper stresses the critical role of class balancing techniques in improving fairness, sensitivity, and reliability in predictive models for stroke detection. It provides actionable insights for their implementation in clinical settings. Keywords: Class Imbalance Techniques, Healthcare Datasets, Machine Learning Models, SMOTE, Stroke Prediction, Synthetic Data Generation. Proceedings Citation Format Fatimah Adamu-Fika, Deborah Ifeoluwa Ayeku, Aisha Tijjani Ramalan, Henry Onyeoma Mafua, Aanuoluwapo Enyojo Baba-Onoja, Tsentob Joy Samson, & Onyinye Vivian Okpoko (2023): Application of Synthetic Minority Over-Sampling Technique to Mitigate Class Imbalance in Stroke Prediction. Proceedings of the 37th iSTEAMS Multidisciplinary Cross-Border Conference. 30th October – 1st November, 2023. Academic City University College, Accra, Ghana. Pp 195-210. dx.doi.org