Load existing Model in Weka Knowledge Flow

Question

I am trying to plot multiple ROC curves in the same diagram in Weka. I have learnt that I can do this in Weka Knowledge Flow using "Model Performance Chart". However, I can't figure out how to do this for existing models.

I have tried using ArffLoader and TestSetMaker to generate the testing data, and connected this to a suitable Classifier icon (eg AdaBoostM1 when this is the kind of model I am trying to load). In the configurations of the Classifier icon I choose "load model" and in the Status bar it says "Loaded model.". However, when I run this it says "ERROR: no trained/loaded classifier to use for prediction".

Can anyone tell me what I am doing wrong here? Thanks in advance!

Answer 1

There is a post that was published here that indicates some ambiguity in the meaning of the error. It also continues to state that the order of attributes and the number and order of values is also rather important.

It also states that 'for performance results to be computed, your Knowledge Flow process will need a "ClassifierPerformanceEvaluator" component after the classifier and before a TextViewer component.'

If you are new with the KnowledgeFlow environment, there is a great tutorial here from Rushdi Shams that details the general process.

Below is a sample workflow that has generated desirable results using AdaBoost (preloaded model):

Hope this Helps!