1
$\begingroup$

Let's suppose a binary classifier with an accuracy of 90%. I wonder if the classifier does not know how to classify the 20% and it is hitting half of them by chance or there is another explanation. Any reference?

EDIT: For this question let's suppose that what we are trying to classify has 50% of samples in each class.

Improve this question
$\endgroup$
2
  • $\begingroup$ 20%.. what does this number represent? Do you mean remaining 10%? $\endgroup$
    – Laksan Nathan
    Commented Mar 2, 2019 at 20:45
  • $\begingroup$ If the classifier does not know how to classify the 20% of the samples it will hit 10% by chance and it will fail the other 10%. The 10% of failures is the error of the classifier. The other 90% is formed by the 10% of samples successful by chance and the remaining 80%. $\endgroup$
    – Daniel Lerch
    Commented Mar 7, 2019 at 14:56

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.