Predictions using Support Vector Machine with Particle Swarm Optimization in Candidates Recipient of Program Keluarga Harapan

Abstract

Program Keluarga Harapan (PKH) is a conditional social assistance program as an effort to alleviate poverty which is allocated to poor vulnerable households. The determination of candidates for the Program Keluarga Harapan assistance recipients is still carried out in village meetings, so it takes quite a long time and there is potential for subjectivity in the assessment carried out by Village Government officials which can lead to differences of opinion between deliberation participants in assessing the eligibility of residents as PKH recipients. For this reason, this research will use an optimization method, namely Particle Swarm Optimization (PSO) to select the most optimal attribute out of 39 attributes. After that, a classification algorithm, namely the Support Vector Machine (SVM), was chosen to form a classification model for Candidates for Social Assistance for the Program Keluarga Harapan (PKH). The classification of Candidates for Social Assistance Recipients of the Program Keluarga Harapan (PKH) was carried out in 2 experiments, namely before and after optimization. Experiments before optimization give an accuracy value of 92.44%. While the Support Vector Machine accuracy value after optimization gives an accuracy value of 92.51%. Based on the experimental results, it can be concluded that the Particle Swarm Optimization method can increase the accuracy of the Support Vector Machine algorithm by 0.07%. And the best model is the Support Vector Machine after optimizing Particle Swarm Optimization by using the 17 most optimized attributes in determining class targets.