ExF-SVM: Exhaustive feature selection with support vector machine algorithm for brain stroke prediction.

Abstract

Predicting brain strokes requires decision-making, and over the past few decades, artificial intelligence (AI) based technologies have greatly improved disease diagnosis. Even with their potential, hospital environments continue to lack trust in these AI models because of their "black box" nature-that is, their inability to be explained or interpreted by medical practitioners. To overcome this gap, explainable AI is emerging, combining techniques that improve interpretability as well as explainability. Brain stroke is one of the most prevalent illnesses that result in death unless proper diagnosis, prediction, and treatment are obtained. Timely and precise prediction of early brain stroke is crucial to preventing additional harm to patients. To alleviate this, advanced learning models use several learning algorithms and approaches for reliably identifying brain stroke. However, the prediction of a brain stroke is not an easy or simple process. Hence, this work proposes a novel feature selection technique for determining the most crucial characteristics and creating an efficient brain stroke risk detection model. To increase prediction accuracy and reliability, this study presents the Exhaustive Feature Selection with Support Vector Machine Algorithm for Brain Stroke Prediction. An exhaustive feature selection-based support vector machine (ExF-SVM) algorithm has been proposed, developed, and assessed in this work for brain stroke prediction. The proposed methodology has been evaluated with the Receiver Operating Characteristics (ROC) curve, sensitivity, specificity, F1-Score, etc. The proposed models' classification results demonstrated the strong influence of improved classification accuracy of 4-14 % compared to the other models and 5-15 % on the F1 score. The results of this work would lead to various innovative contributions and useful ramifications in healthcare.