As the title, after I performed a Feature Selection, is it mandatory to respect the same ratio (between development set and test set) in Model Selection?
Feature selection has nothing to do with a model. It is just finding a candidate list of related (or even unrelated) features that might be in a model, so it this put there is no rule that says it has to be from train. It could be from another similar model that someone else created. That actually gives it better validity since it was already determined as a valid feature in a similar model. However once the features are made part of the training set, they should be in the test set.
If my understanding is right, You have selected some set of features using any of the feature selection technique and going to train(develop) the model.
Now you want to know is like whether i have to use the same set of features in the test set as well.
If so, it is definitely YES.
Actually for the whole data itself we apply the feature selection process and then only we split the data for develop(train) and test.
Note: If my understanding is wrong or you need something else, please give more input.
Thanks
Geeks Mental is a community that publishes articles and tutorials about Web, Android, Data Science, new techniques and Linux security.