0
$\begingroup$

I am using Random forest (Matlab) to classify the binary data. Broadly, the input to the random forest is number of features and class label. And random forest, after training, return the labels for "unseen" data (30% of the overall data). I have two (very basic) questions related to performance evaluation:

  1. I don't have any threshold, I am using simple random forest implementation in Matlab. Is there any way to compute ROC curve? Will the threshold be "number of tress"? (I am very new to ROC curves, just read about these)

  2. I already calculated TPR, FPR, F-scores and confusion matrix, are there any other performance evaluation measures for binary classification? Pointers or just names will be good help.

Suggestions/Comments?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

  1. When you say that you don't have any threshold, do you mean that you only get predicted class labels, but not classification scores? Without scores, you cannot compute a ROC curve. If you do get scores, take a look at the perfcurve function in the Statistics and Machine Learning Toolbox.

  2. The doc for the perfcurve function shows other evaluation measures.

Improve this answer
$\endgroup$
0

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.