Building high-precision classifier with RapidMiner

leonp

Supposedly, I need to build classifier with high precision and low recall. Which operator should I use in RapidMiner? For example, ApplyModel operator doesn't support any intervention in threshold? It's also unclear how the same ApplyModel operator works with models of different modeling: NeuralNetworks, SVN, etc.

MartinLiebig

Hi leonp,

 

Apply model creates (in case of binary classification) three new attributes, the prediction and two confidences. If you want to optimize on the threshold you can use the dedicated operators like Create Threshold, Apply Threshold. There are also further operators working on the confidences e.g. Select Recall and Find Threshold.

 

~Martin