TPR@FPR

rvaz

Hi,

Does anyone know how can I get the True Positive Rate at a specific False Positive Rate (TPR@FPR) in a Binominal Classification?

Thanks in advance,

Rui Vaz

land

Hi,
if I understood you correctly, the classification of one example set will give you exactly one TPR, FPR pair, isn't it? So you will have to classify the same data set many times with varying learning parameters, or which way do you want to get more than one FPR at all?

Greetings,
  Sebastian

rvaz

Hey Sebastian Land thanks for the reply,

You're correct the classification will give me a TPR and FPR.

But since TPR(y-axis) varies with FPR (x-axis) in the ROC curve,
I was wondering if there's a way of getting the value of TPR at a specific FPR.

I need that value as a fitness for my implementation in java.
                     
I hope I was clear.

Thanks,
Rui Vaz

land

Hi,
there's actually already a feature request for an operator, that transforms a roc curve back to an example set.

Since it will take quite some time until we manage to implement that, you can otherwise transform your data manually using a complex rapid miner process or write the operator yourself.

Greetings,
  Sebastian

rvaz

Ok Sebastian I'll work on it.

Thanks,
Rui Vaz

land

Hi,
if you have created such a process, you might upload it with the new community extension, so that it can be used by other community members. Because it would be a nice example and would save a huge effort for everybody that would have to rebuilt it

Greetings,
  Sebastian