IMPROVING HEART DISEASE PREDICTION THROUGH PSO-OPTIMIZED SUPPORT VECTOR MACHINES USING LINEAR AND GAUSSIAN KERNELS

This study examines the performance of Support Vector Machines (SVMs) utilizing two distinct kernels, linear and Gaussian, optimized through two variants of the Particle Swarm Optimization (PSO) algorithm: classical and adaptive. The focus is on classifying heart disease data, which is crucial in medical diagnostics where precision and efficiency are vital. PSO, an evolutionary optimization technique, is employed to fine-tune SVM parameters, thereby enhancing the models' classification accuracy. The heart disease dataset used includes multiple features representing various health indicators, necessitating a robust classification methodology. Performance is evaluated using accuracy, precision, recall, and F1-score metrics. The results reveal how different PSO variants influence SVM classification, providing a comparative analysis between the linear and Gaussian kernels. The findings indicate that PSO-optimized SVMs can substantially improve heart disease detection and diagnosis, with each kernel demonstrating unique strengths depending on the optimization strategy employed.