MACHINE LEARNING-BASED EARLY STROKE PREDICTION USING CONVOLUTIONAL NEURAL NETWORKS

Abstract

              A brain stroke is mostly brought on by modifications in the blood flow to certain brain regions. The patient's quality of life may consequently be diminished as a result of some specific functions associated with the affected area being restricted.As a result, machine learningbased prediction approaches used in clinical decision, including anticipating the onset and course of diseases as well as assisting physicians for medication.This approach to stroke predictive analytics technology was implemented using machine learning models on brain disease datasets.  The proposed of this model is to create a machine learning application to identify stroke.Low prediction accuracy, precision, and recall, as well as high temporal complexity and detection error rate, are some of the disadvantages of ML approaches.  To address these problems, we suggest a novel approach called the machine learning-based Convolutional Neural Network (CNN) model, which can accurately and efficiently predict a patient's risk of stroke.  Brain illness datasets were used to test machine learning models for this predictive stroke analysis technique.This model aims to develop a machine learning application that uses CNN to identify strokes. This model, which is an ad hoc version of a multilayer perceptron, is employed in feature selection to determine an attribute's maximum threshold. Because it is a healthcare dataset, the predictive model only uses 11 features and one target class. Consequently, in order to extract the features that have the using feature selection methods. With a 92.5% accuracy rate, the model surpassed other machine learning models in our comparison of their accuracy.