AI-Based Cerebral Vascular Accident (CVA) Analysis and Prediction

Abstract

Early stroke detection significantly increases the prognosis for both survival and rehabilitation. Patients are more likely to receive appropriate therapy that minimizes brain damage and lowers the risk of consequences if a stroke is detected early on. Researchers are motivated to investigate the possibilities of artificial intelligence and machine learning technologies in creating new categorization systems that can identify and detect strokes more quickly and accurately due to their rapid development. This could potentially enhance the likelihood of surviving and recuperating. The support-vector machine (SVM), logistic regression, decision tree, random forest, Bayes nets, and K-nearest neighbor (KNN) algorithms are employed in this study's CRISP model technique. To enhance the final quality, the dataset was balanced using an oversampling technique, and the algorithms employed were subjected to principal components analysis (PCA). With an accuracy rate of 99%, the Random Forest algorithm is regarded as the optimum for prediction. Our study illustrates that the random forest classification model using the data balancing strategy outperforms the other strategies investigated, with a 99% classification accuracy and a 98% F1 score. The study also shows that the outcomes are unaffected by preprocessing with the PCA technique. The next objectives of the study are to use a larger dataset, various preprocessing methods, and machine learning models to enhance the framework models.