Problem with generating ROC Curves for Anomaly Detection extension.

Question

I am trying the Generate ROC operator after running different k-nn algorithms in the Anomaly Detection extension.  However, I am getting a error message telling me " not found in labels.  All worked fine when I was just using the k-nn algorithms.  But when I added the "Generate ROC" operator in series after the k-nn, I got the error message.  I am a newbie running RapidMiner 5.3.013.  Is it how to generate the ROC curve?  Also, how do I add different k-nn algorithms output to generate different ROC curves for each algorithm on the same graph to compare them.  Many thanks!

Frank

liming · Answer

Hi
  My dataset is pen-global-unsupervised-ad3.csv. And my process XML is test2.xml. I want to get a ROC curve.

pen-global-unsupervised-ad3.csv

test2.xml

Telcontar120 · Answer

Can you post an xml with your process and a description of the specific error you face?  The OP is from a version of RapidMiner that is 5 years old!  I am not experiencing any problems with the Performance operator and generating the ROC curve, but to do so you need to formulate your problem as supervised machine learning, i.e., you need a label for it to compute performance.  If you are simply using the outlier classification unsupervised approach then the ROC curve would be undefined in such a case.

liming · Answer

Hi,
  Did you get the answer? I want to use the anomaly detection extension in RapidMiner and get the ROC curve for different algorithms? I would be grateful if you could solve my problem.