An Ensemble Learning Approach and Analysis for Stroke Prediction Dataset

A stroke is an illness that results in traumatic brain injury by tearing blood vessels. A brain stroke can also occur if blood flow and other nutrients to the brain are interrupted abruptly. It is one of the major global causes of disability and death, as per the report given by the World Health Organization (WHO). With increased convergence amongst technology and medical diagnosis, practitioners create possibilities for improved management of patients by comprehensively quarrying as well as archiving patient's records containing their medical background. As a result, it becomes critical to investigate the interdependence of these factors (risk) in patient's medical records and comprehend the relative impact of these factors for the prediction of brain stroke. This research establishes an early estimation of stroke diseases by combining the existence of hypertension, heart disease, body mass index, smoking status, prior stroke, age, and some other feature attributes. For forecasting the stroke, various statistical methods and five different ML models including some ensemble learning techniques like Support Vector Machine (SVM), Random Forest (RF), Ada-Boost Classifier (ABC), Decision Tree Classifier (DTC), and XG-Boost Classifier (XGB) were used to train the feature attributes. Furthermore, the proposed research work has accomplished an accuracy of 95.08 percent, with the XG-Boost Classifier outperforming the Machine Learning (ML) Models. As a result, XG-Boost is nearly the most preferable classifier for predicting strokes, which can be used as a reference model by physicians and also used by patients considering aid in the early detection of a potential stroke.