Diagnosis of Heart Disease Using a Novel Membership Computation Method Within a Fuzzy SVM Framework

Heart disease ranks among the most prevalent diseases globally. This study aims to improve heart disease diagnosis by utilizing a fuzzy support vector machine (SVM). Traditional fuzzy SVMs have effectively minimized the impact of outliers by assigning appropriate fuzzy memberships to each data point. Nevertheless, these models often overlook how outliers influence the positioning of class centers. In response to this issue, our paper introduced a novel fuzzy membership function which was applied to both linear and nonlinear fuzzy SVM. This approach identified outliers before establishing class centers. Additionally, samples positioned near the hyperplane were assigned increased decision weights. Such a strategy significantly reduced the influence of outliers on class centers, thereby boosting the effectiveness of the fuzzy SVM. The experimental results validated the efficacy of our method. The tailored approach not only mitigated the impact of outliers but also markedly enhanced classification accuracy and generalization capabilities. This confirmed the contribution of our enhanced fuzzy SVM to the field of heart disease diagnosis.