Peningkatan Akurasi Metode Support Vector Machine melalui Particle Swarm Optimization pada Penyakit Ginjal Kronis

Abstrak

Abstract: Chronic kidney disease has a significant impact on death rates from disease and continues to increase worldwide. Therefore, this research was carried out with the aim of being able to detect the disease early before it is diagnosed at a more serious stage. The application of the individual SVM method and the PSO-based SVM method was carried out to find out which model gave the best results in detecting chronic kidney disease. The SVM method has advantages in accuracy but has difficulty selecting features for the best input. So PSO Optimization (Particle Swarm Optimization) was chosen to overcome the weaknesses of SVM, because the PSO algorithm has advantages in selecting features and providing attribute weights. The results shown are that the PSO-based SVM method is successful in improving model performance and increasing the weights on attributes. From the evaluation results, the SVM kernel Dot is superior to other kernels, as proven by the SVM method alone, the accuracy was 92.25%, but after optimization using PSO, the accuracy value increased to 99.50% so there was an increase of 7.25%. Conclusion that the PSO optimization algorithm can increase the effectiveness of SVM performance, in order to predict kidney disease.

Keywords: chronic kidney, prediction, svm, optimization, pso