How to Split data into 3 equal parts (A,B,C) with balanced numbers of each subclass.

Hi，

I have a dataset and I can split it into 3 parts but I do not know how to make sure that these 3 parts are with balanced numbers of each subclass.

Answer 1

[the cyclist]

Probably the easiest way to do this would be with the following steps:

1. Isolate each of the classes, using something like "Y == A" (this is pseudocode, not MATLAB syntax), etc
2. Randomly assign each class to your groups (training and testing, whatever)
3. Combine the classes into the groups

If you upload the data -- or at the very least describe in detail what your data look like -- I could probably give more specific advice with actual code.

[Xue Zhang]

Thanks,that's my first idea too. The data is 2400 images with 24 subclass. And I also need to do the cross validation to these.