I want to create fitness function to be used in Genetic Algorithm for Feature Selection and those features are used by SVM for classification. How to create such a fitness function?
Thanks for your question. I think it is better if I go for a detailed answer rather than an objective one for your concern. Hence, I went for a search and found few valuable files for you. I think you can get a detailed answer of your concern if you just give some time to these files. Please feel free to ask more if you have any doubts in these files.
Thank you so much your your detailed and relevant answer. I am studying the papers being recommended by you and will hopefully figure out the problem. I really appreciate your help.
The most straightforward fitness function in your case would be the classification performance of an SVM model built on the selected features on a separate test set.
Dear Davide Roverso, theoretically I understand that and I know that recall rate will be the performance measure, but I am feeling it difficult to create/implement such a function in matlab.
Sorry Haris, but I cannot help on detailed programming issues. I moved away from Matlab 3 years ago and I'm not sure how fitness functions are defined in the GA toolbox .... but I guess you will have to define a function that takes in input the feature set to evaluate, then calls the SVM training (with some default parameters), and finally evaluates the obtained model on a test set.