Stroke Detection Through Ensemble Learning: A Stacking Approach

A stroke is a health condition that occurs when there is an interruption in the normal blood flow to the brain. Brain cells eventually die due to the lack of blood supply. As per the WHO, it ranks as the fundamental cause of both fatalities and disabilities. Timely identification and acknowledgment of symptoms can swift stroke treatment, leading to enhanced health outcomes. In this work we developed different machine learning models including stacking. The selected models are Random Forest (RF), Decision Tree (DT), Logistic Regression (LR), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM). The base models used for stacking are DT, LR and KNN and for meta model LR is used. The dataset was preprocessed by deleting unnecessary columns, dropping null values, removing outliers and by encoding the categorical values using One Hot Encoding. Oversampling technique is used to balance the dataset as it was highly unbalanced. After evaluation the findings reveal that the proposed stacking ensemble approach performs better than the single models, achieving an accuracy rate of nearly 99%.