[SOLVED] Bayes NET

studentstudent Member Posts: 17 Contributor II
Hello,

I would do the following:

Input: a,b,c,d,z (random variable/ table columns)

1. Create a decision tree which shows the influences on z

Example: http://monkpublic.library.illinois.edu/monkmiddleware/public/analytics/decision_tree.png

2. Create a probability table for z.

Example:
z : P(z|a,b,c,d)

z_1 : 25% with a=1,b=1;c=0; d=true
z_2 : 12.5% with a=1,b=1;c=0;d=false
etc.

My problems with the bayes.NET are

a) The probability table is very large, can I write directly for z in a file (Bayes.NET gives me unfortunately for each variable, a table...).

b) I can learn the tree-structure restricted to z?


mfg student

Answers

  • MariusHelfMariusHelf RapidMiner Certified Expert, Member Posts: 1,869 Unicorn
    Hi,

    concerning the tree structure, did you watch the introductory videos on our website? In general, you can use one of the Decision Tree operators to create tree structures.

    Best regards,
    Marius
  • studentstudent Member Posts: 17 Contributor II
    Thank you. but I see just the real problem: I have an out-of-memory-exeption.

    My questions are:

    1. Can you tell me how I can summarize table row?
    Example:
    [id1,ja,100,50]
    +
    [id1,ja,50,12]
    =
    [id1,ja,150,62]

    2. I would like a "loop value" loop which I only create the line which outputs in a special column "y" has the value "1".

    "loop value" with "attribute=y" and "iteration macro =1" gives me the columns with 0 and 1 => out of memory


    mfg
    student
  • studentstudent Member Posts: 17 Contributor II
    hi,

    the functions "aggregate" and "filterExamples" solve my problem.

    how I can save one attributs probability table (from bayesNet) in a file?
  • MariusHelfMariusHelf RapidMiner Certified Expert, Member Posts: 1,869 Unicorn
    Which bayes net operator exactly are you using?
  • studentstudent Member Posts: 17 Contributor II
    hallo,

    weka.classifiers.bayes.net.search.local.K2  -- -P 22 -S BAYES

    and


    weka.classifiers.bayes.net.estimate.SimpleEstimator -- -A 0.5

    My problem is that the probability tables for some nodes are not displayed. (it would probably be too big, I have 13 gb RaM)
  • MariusHelfMariusHelf RapidMiner Certified Expert, Member Posts: 1,869 Unicorn
    So obviously you are using the W-BayesNet operator.
    How many examples are you using? Can you please post your process setup? Or even better, can you create a process which demonstrates the problem which uses only the samples data set or the Generate Data operators, such that we are able to reproduce the problem?

    Furthermore, let me ask which version of RapidMiner and the Weka extension are you using? The latest version of RapidMiner is 5.3.7, of the Weka extension 5.3.1. If you are using lower versions, please upgrade.

    Best regards,
    Marius
  • studentstudent Member Posts: 17 Contributor II
    thanks. I have found my error.
Sign In or Register to comment.