2
$\begingroup$

When I evaluted my model, I always get equal values for metrics : accuracy, precision, recall, and f-measure. Is that possible?

$\endgroup$
1
  • 3
    $\begingroup$ Beware; these are all improper accuracy scoring rules, as discussed here. $\endgroup$ Commented Jun 29, 2019 at 11:28

1 Answer 1

1
$\begingroup$

Yes, possible, first of all, everything can be $0$ or $1$, and so equal. In other cases, if precision and recall are equal, then F-measure is equal to them, because it is the harmonic mean of the two. Let's contemplate about how the first three can be equal:

$$\text{Precision}=\frac{TP}{TP+FP}, \ \ \ \text{Recall}=\frac{TP}{TP+FN}, \ \ \text{Accuracy}=\frac{TP+TN}{TP+TN+FP+FN}$$ If $FP=FN$, then precision is equal to recall. Then, $$\frac{TP+TN}{TP+TN+2FP}=\frac{TP}{TP+FP}\rightarrow (TP+TN)\times FP=2TP\times FP$$ So, if $FP=0$, i.e. everything is correct; or, if $FP=FN\neq 0$ and $TN=TP$, all three (consequently all four) are equal.

$\endgroup$

Your Answer

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

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