Optimasi Kernel SVM dengan PSO untuk Gagal Jantung

Abstract

Accurate early detection is important to improve the quality of life of patients and reduce mortality and a major burden on the public health system caused by heart failure. This study aims to improve the accuracy of heart failure prediction using Support Vector Machine (SVM). SVM is used as a strong classifier for high-dimensional data, then optimizes its kernel using Particle Swarm Optimization (PSO), which has not been widely applied in similar studies. The method used includes computational experiments with a quantitative approach based on heart failure datasets from the UCI Repository which are analyzed using SVM with three types of kernels: Dot, Radial, and Polynomial. PSO is used to optimize the selection of kernel parameters in SVM to improve classification accuracy. The results show that SVM + PSO kernel Dot gives the best performance, with an AUC of 0.865 and an accuracy of 83.97%, and this difference is confirmed significant through a paired t-test (p <0.05) compared to SVM without optimization. PSO optimization consistently improves precision and recall in the tested kernels, indicating stability and effectiveness in classification. The impact of the research is to make a significant contribution to early detection efforts for heart failure which can lead to faster treatment and improved quality of life for patients, but also adds clinical value for medical practitioners seeking efficient and accurate classification methods.