Improved Accuracy in Early Identification of Ischaemic Stroke using K- Nearest Neighbors with Support Vector Machine

Abstract

The major aim of this study was to use MRI scans as a diagnostic tool for identifying strokes in the brain. K-Nearest Neighbors, an innovative alternative to the Support Vector Machine, was used to improve accuracy and specificity beyond what had been achieved before. K-Nearest Neighbors (with a total of 20 participants) and Support Vector Machines (with a total of 10 participants) are compared and contrasted here. (which had a total of 10 participants). Alpha = 0.05, the enrollment ratio = 0.1, 95% confidence interval, and pre-test power = 98% were used in conjunction with the G Power software to arrive at the final sample size. In comparison to the Support Vector Machine algorithm's 89% accuracy and 76% specificity, the unique K-Nearest Neighbors algorithm achieves 97% accuracy and 89% specificity using the proposed method. According to the data, the level of statistical significance attained for accuracy is p = 0.005, while the level of significance attained for specificity is p = 0.045. These results are provided in light of the research's conclusions. When comparing K-Nearest Neighbors to Support Vector Machine classifiers, the state-of-the-art K-Nearest Neighbors method outperformed its predecessors.