0
$\begingroup$

I'm trying to predict the gender of a Twitter account using only the profile information like tweet text, description and used colors.
I've trained a SVM classifier and then tested dividing the initial data set in two portions (80% training - 20% testing).

After that, i would like calculate accuracy and other metrics like precision and recall.
My doubt here, is how should I calculate precision and recall? I mean, from my point of view it's more correct calculate them for both classes (male and female).
It's like a binary problem but the two classes are equally important during the prediction phase.

What do you think?

$\endgroup$
0
$\begingroup$

First - this is not a multiclass problem. With only 2 classes that is the definition of not-multclass. It is a binary problem

Recall and precision are defined when you have one class that you call positive (usually because it is the "most" important class). In your case, both classes are "equally important". In this case you can calculate the global accuracy - as you did - and calculate the accuracy for each class (# true predictions for class M/ # examples of class M) and the same for class F. Recall is the accuracy of the positive class.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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