8 views (last 30 days)
I have not run across anything with the Classify function that indicates the number of observations for each group must be equivalent. This has been done previously using LDA (ie Altman z-score paper) but I just want to be clear it is not required using this function.
Veera Kanmani on 20 Apr 2018
Kenta on 31 Mar 2020
For example, in Support Vector Machine (SVM), you can specify the weitht of each class by defining some parameters such as cost and prior. I am not sure if my answer can response to what you intended to ask, but I hope it can help you.
If you impose a great penalty if A is classified into B, such inference will not happen a lot. When the number of images (sample) which belong to class A is much greater, many test images might be classified into A. In that case, you can weight each class.