Current Proceedings on Technology

Yazarlar: Sanghun Yun, Won-Seok Kang, Hyeong-Oh Kwon

Konular:-

Anahtar Kelimeler:Electrocardiogram,Classification,Support vector machine,Particle swarm optimization.

Özet: Electrocardiogram (ECG) is a biological signal and can provide the information of human heart status. Many heart diseases can be found by analyzing ECG. Thus ECG analyzing method with good performance is very helpful for determining human heart status. Support vector machine (SVM) is a popular pattern classification method in ECG analysis and diagnosis. Kernel parameter setting in the SVM training procedure significantly influences the classification accuracy. Therefore, it is necessary to develop an automated and reliable approach to determine the values of these parameters. In this paper, we presents the SVM parameter optimization approach using Particle swarm optimization (PSO). An important factor that influences the performance of SVM is selection of parameters. The SVM is based on multiclass classification (C-SVM) and we experimentally select 5 parameters for optimization: (kernel function: t, cost value: c, gamma value: g, degree value: d, coefficient value: r). To estimate the performance of parameter combination, we estimate the classification accuracy of each parametercombination and used 6-cross validation to estimate the accuracy of each parameter combination. Our method attempts to increase the classification accuracy. From experiment results, it can be concluded that our approach can be solve the problem of estimating the optimal SVM parameter settings.

---