To get High accuracy by leaving low precision classes

opusminer

Hi!!

I am using RapidMiner Studio 6.0.008.

To give some background I am trying to predict the posting Accounts for a given invoice using decision tree. (If you don't understand it doesn't really matter)

I used X-validation and as I can see from the PerformanceVector  I got 87.06 % accuracy. I am ok with the model and I want to go use it.

However, since the application is very sensitive I want to use the cases where I have 100% class precision.

What I want is when I give my model a test example(unseen example) the model should :
      - predict only when it is 95% sure (to make it more general say above some threshold )

opusminer

I found the operator Drop Uncertain Predictions.