How to pass parameters to the W-BayesNet operator?

gladysCJ
gladysCJ New Altair Community Member
edited November 5 in Community Q&A
Hi:

First, congratulations to all of you for creating the Rapid-I forum and specially for such an awesome program.

My problem is the following.  I'm trying to use the Weka classes for Bayes Nets in RapidMiner but I have some problems for getting the operator  W-BayesNet  to work as I need. The problem is that I  need to pass parameters  to the weka classes but so far I  have not been able to find the best way to do it. 

For instance, to implement the K2 algorithm, the Q parameter of W-BayesNet  is set to weka.classifiers.bayes.net.search.local.K2. But it is not enough because I need to pass some parameters to this Weka class, as for instance,  the parameter  P  (maximum number of parents), S (score), etc.  ( In Weka we use :  -Q weka.classifiers.bayes.net.search.local.K2 -- -P 3 -S BAYES) 

Any suggestions are very very welcome!!!

Thanks in advance,

Gladys
Tagged:

Answers

  • IngoRM
    IngoRM New Altair Community Member
    Hi Gladys,

    we will have a look into this and write back if we find out how this can be done...

    Cheers,
    Ingo
  • gladysCJ
    gladysCJ New Altair Community Member
    Hi Ingo:
    Have you find a way to pass the parameters to the W-BayesNet class? The problem is that next week I will teach a graduate intensive course on Bayes Net for Data Mining and  I would like that the students could use RapidMiner instead Weka  for evaluating several learning algorithms for BNCs.
      Thank you for all your support.
      Cheers,
        Gladys
  • IngoRM
    IngoRM New Altair Community Member
    Hi Gladys,

    I have good news: I found the error. It was actually produced by a bug in Weka for the option handling for the BayesNet learner. It only worked correctly when -E option was also set. I fixed the bug in Weka and changed a setting in RapidMiner (which accidentally removed the -E option due to this Weka bug) and now everything works fine for me. I will send a bug report to the guys in Waikato so that they can fix the error for future versions.

    For now, you can access the fixed version (including a fixed Weka library) via CVS. Please refer to http://rapid-i.com/content/view/25/48/ for a description how this can be done. Of course this fix will also be part of the next release.

    Cheers and thanks for pointing this out,
    Ingo

Categories