PSO-SVM Classification technique

Hi

I am trying to use a PSO-SVM classifier for classification in my project. Going thru a lot of research papers I could not figure out 1 thing that how exactly do you predict 'C' n 'Gamma', parameters for SVM using PSO. by reading the papers it looked as if they just applied it using hit and trail method. No paper explains the theory behind linking the output of PSO i.e the gbest n pbest or the best velocity and position to parameters of SVM .

Plzz Help Thanks

Answer 1

[S. Muhammad Hossein Mousavi]

Hi

I just write a code for combining Neural Networks and PSO for classification task:

https://www.mathworks.com/matlabcentral/fileexchange/104350-pso-classification-algorithm

Hope it helps :)

Answer 2

[kh rezaee]

For better classification, I suggest you use PSO to feature selection and SVM to classify data. But combination of PSO and SVM is little difficult.

[Pankaj Kumar Behera]

can i get some sample matlab codes using which i can optimise c and gamma parameters of a multiclass svm

Answer 3

[Dhemiebabs]

Any update on how you were able to solve this(PSO-SVM) please. I'm presently carrying out the same project and I'm stock at the same point.

I've been on it for months and I'm to present it next month.

Your prompt response is much appreciated.

Thanks.