Weights in LibSVM to get balanced sample

In777In777 Member Posts: 29 Contributor II
edited November 2018 in Help

E.g. I have a binary classifier with 3000 positive instances (sentences) and 12000 negative instances. To get balance sample I want to use class weights paramter of LibSVM (linear SVM). Do I define weights as 0.75 for positive and 0.25 for negative class or 0.5/0.5, hence making them of equal importance?

Sign In or Register to comment.