I'm looking into some classification tasks at the moment. The test data is unbalanced where one particular class is half the data and the remaining 5 take up the remainder of the test data. When I look at the output for some of the classifiers I notice that the AUC is over .8 while the TPR and FPR might be something such as 0.4 and 0.0234 respectively.
So can anyone explain to me what might be going on here? I'm using the WEKA API so I'm assuming there evaluation code is fine. I'd of thought that a higher TPR would get me further along with a stronger AUC and that a lower TPR would return a lower AUC.
