Optimasi Algoritma Support Vector Machine Menggunakan Particle Swarm Optimization dengan Kernel RBF, Linear, Polynomial, Sigmoid dalam Klasifikasi Data Stunting

Abstract

In the development of health in Indonesia, stunting is a severe problem in health and welfare. The problem that often arises at the Medan City Health Office is when the data collected for stunting evaluation is consistently inaccurate and uncertain every month because only estimates are calculated based on cases carried out by the puskesmas before. Support Vector Machine is the most powerful and newest technique in the prediction that can be used as a numerical prediction and classification. However, the Support Vector Machine still needs to improve in selecting appropriate and optimal features for the attribute weights, causing the prediction accuracy to be low. Not all data can be separated linearly, while Support Vector Machine can only sell data linearly, so development is needed to make Support Vector Machine able to combine non-linear data, one of which is by adding a kernel function. This research determines the best classification using the Support Vector Machine algorithm based on Particle Swarm Optimization on Radial Basic Function, Linear, Poly, and Sigmoid kernels. The best performance result in increasing accuracy is the Radial Basic Function kernel.