I've two sets one is ground truth and other is output of my machine learning models. Assume my groundtruth set is A={1,2,3,4,5} and output of machine learning model is B={3,4,5,6,7,8}. One way I can measure performance is IoU(Inersection over union) like object detection. So IoU will be

A intersecton B/ A U B

I'm not sure is it reasonable? Or any other better evaluation metric can I use to measure the performance?


"I'm not sure is it reasonable?"

yes it is. The metric you refer to is known as the Jaccard index.

| improve this answer | |
  • $\begingroup$ Is there any evaluation metric? $\endgroup$ – Shahrear Bin Amin Mar 25 at 10:24
  • $\begingroup$ there are other metrics you can think of, for instance a variation of the Jaccard index called overlap coefficient, en.wikipedia.org/wiki/Overlap_coefficient. But at the end you should think which metric aligns which your goals. $\endgroup$ – Alejandro Celis Mar 25 at 14:27

Not the answer you're looking for? Browse other questions tagged or ask your own question.