Prediksi Kelulusan Mahasiswa Menggunakan Support Vector Machine Berbasis Particle Swarm Optimization

Abstract

One of the criteria for a good accreditation grade must be to meet the student's timely graduation rate. The punctuality of students is still a crucial problem in several universities in Indonesia. Many factors cause students to fail to complete their studies on time. Student data owned by the university can be processed using data mining techniques to become useful information for this graduation problem. Based on these problems, research was conducted for student graduation using the Support Vector Machine (SVM) based on Particle Swarm Optimization (SVM+PSO). The test results obtained in this study are the linear SVM Kernel Model Accuracy of 95.91%, the RBF Kernel SVM Model Accuracy of 95.91% and the Polynomial SVM Kernel Model Accuracy of 96.20%. Meanwhile, the SVM+PSO Merger Model increased the Linear Kernel's accuracy to 96.12% (+ 0.21) and the RBF Kernel to 96.24% (+ 0.33) but experienced a decrease in the accuracy of the Polynomial Kernel to 96.12%. (- 0.08).