Extension Weka - Predictive - W-BayesNet (TAN): problem in export WekaResult model

ferreiradossant
ferreiradossant New Altair Community Member
edited November 5 in Community Q&A

Hi.

So, I have a data set that I used to construct my TAN model.

Now that I have the model I want it to be applied in the scoring of another data set with missing models. 

The problem is that I need to store the model (Weka result) and I can´t.

I used the "write PMML" operator and it gives me an error message: "The class Wekaclassifier cannot be exported to PMML. For some classes are no PMML writes available due to restrictions of the standard or other technical reasons".

I also try the "Write model" operator, storing it as XML file, but it also does not work.

The problem is that I want to open the model in the SamIam software. It only read hugin net.

I search everywhere and I cannot find a solution to my problem.

Could you please help? I attach a picture of the process.Captura de ecrã 2017-07-18, às 16.49.12.png

Best Answer

  • Thomas_Ott
    Thomas_Ott New Altair Community Member
    Answer ✓

    The PMML write operator is limited in the types of PMML it can export too. As far as I know, there is no workaround to get this to import into Samlam(?).



Answers

  • Thomas_Ott
    Thomas_Ott New Altair Community Member
    Answer ✓

    The PMML write operator is limited in the types of PMML it can export too. As far as I know, there is no workaround to get this to import into Samlam(?).



  • Telcontar120
    Telcontar120 New Altair Community Member

    Setting aside the PMML operator, which has a lot of limitations (not unique to RapidMiner but inherent in the PMML protocol) it should still be possible to write WEKA models using the normal Store operator.  However, there have been multiple reports in the forums of people experiencing problems viewing or writing WEKA models.  I've had similar problems myself.  Can you post a sample process using only Store and the associated error?  It would be really great to get some more examples to the technical support staff so this could be fixed in an upcoming release since WEKA models are quite popular in RapidMiner.

     

  • ferreiradossant
    ferreiradossant New Altair Community Member

    Hi, Thomas.

     

    The solution that I found (after spending an entire afternoon on it) was to apply a "Write ARFF" after the cross validation, extracting the example set. After, I opened Weka software, import the ARFF file, chose all the variables and click on visualize the graph.  To finished, I saved the graph in Weka, as a BIF file, and open it in SamIam.

     

    It took time but I got there. 

     

    Thanks for trying to help me.

     

  • ferreiradossant
    ferreiradossant New Altair Community Member

    Hi, Brian.

     

    I am sorry but I did not understand what you need me to do.

     

    I already write how I solved it. If you have any questions, please let me know.