2

I already referred the posts here and here but they are different and doesn't answer my problem.

I am currently working on a binary classification problem with random forests, neural networks etc.

During assessment of the model, I generated the below KS-statistic

enter image description here

While I understand that KS-statistic indicates the seperation power between both classes,

a) what does it mean to high seperation in Decile 1 vs 2 vs 5 or 8? and What is a good value for KS-statistic?

The Great
  • 3,272
  • So you compute a KS statistic for each model and then want to use this to decide which model is the best? – frank Mar 24 '22 at 09:53
  • @frank - Doesn't it make sense to use KS-statistic for a single model? I built lift, gain and KS-statistic for a single model. No multiple model comparsion. As my dataset is imbalanced, I thought of using these approaches to assess my model performance – The Great Mar 24 '22 at 09:57
  • This is the related post - if you want more info about my problem - https://stats.stackexchange.com/questions/567964/discriminatory-model-but-no-discriminatory-features – The Great Mar 24 '22 at 10:01

0 Answers0