I am building some model, which predict on basis of highest probability from history and I am assuming this is best action. I am comparing this with real action.

Predicted : process resumed somedate2

Actual : process resumed at somedate1

As both are similar and assumed this is correct prediction.

I have applied cosine similarity for same and doing comparison.

Is there any better way to compare same and to check accuracy ?

Any help will be appreciated.

  • $\begingroup$ Its not clear to me what problem you are trying to solve and what you actually doing. Can you please try explaining this again? + adding relevant code/data samples $\endgroup$ – yoav_aaa Mar 4 at 9:06
  • $\begingroup$ It seems like you don't need confusion matrix. CM is an evaluation tool for task of classification, where targets are defined and don't change. The task you are working on seems more like generating it's own output and it's problematic for CM. Cosine similarity definitely seems like a good idea and best solution I see so far. $\endgroup$ – chacid Mar 4 at 9:27

You could create 2 new columns (1 for the truth, 1 for the prediction) which have a more generalised value like PROCESS_RESUMED. You can then use those columns to create the confusion matrix.

  • $\begingroup$ brilliant, did it solve your problem? can you accept the answer if it did? thanks! $\endgroup$ – Jeremie Mar 6 at 14:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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