r - Odd cross validation results in text classification (RTextTools) -
i new text classification. using r package rtexttools perform 'simple' text classification task. cross validation results little bit odd me:
especially accuracy values > 1 baffle me. me, since accuracy := (tp + tn)/(tp + tn + fp + fn), accuracy should not larger 1. values < 0.02 seem little bit off compared other results got. looked source code of r package , found formula package uses compute accuracy:
cv_accuracy[i] <- recall_accuracy(allcodes[rand==i],pred)
based on name accuracy of recalled values computed. me, still not explain values got.
thanks help!
cheers
Comments
Post a Comment