This question already has an answer here:

Are there general rule of thumbs for acceptable accuracy, precision, sensitivity, and specificity values/thresholds in classification? I would imagine that this depends on different applications. I would appreciate it if someone can give me a general guideline.


marked as duplicate by Stephan Kolassa, mkt, mdewey, Siong Thye Goh, Peter Flom Apr 11 at 13:09

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • $\begingroup$ As you already assumed, it highly depends on the problem at hand and its bayes error rate. $\endgroup$ – deemel Apr 10 at 18:46
  • 1
    $\begingroup$ No, there isn't. The proposed duplicate treats the question of how to know whether we have reached the end of the tether in a ML problem, performance-wise, which is just another way of asking whether the current performance is acceptable, or whether it "should" be possible to improve on it. $\endgroup$ – Stephan Kolassa Apr 10 at 20:30