3 ms·
No, accuracy is the same here it is for any binary classification. accuracy = (true positives + true negatives) / all samples. A largely useless measure. As a s
by notascientist 7y ago
No, accuracy is the same here it is for any binary classification. accuracy = (true positives + true negatives) / all samples. A largely useless measure. As a single score F1 score is much better, but sensitivity and specifity are specified separately for good reason. It makes napking math much easier, for one thing.
- skissane 7y agoWhat confused me is there are actually two definitions of accuracy. Accuracy: ACC = (TP+TN)/(P+N) Balanced Accuracy: BA = (TPR+TNR)/2 The definition I gave is the second not the first. You are probably correct that the conference abstract I linked to is using it in the first sense not the second. (I'm not 100% sure though.) Source: https://en.wikipedia.org/wiki/Template:Confusion_matrix_terms https://en.wikipedia.org/wiki/Template:Confusion_matrix_term...