Can't generate decision tree from DAN - NullPointer

Issue #348 new
Iago París created an issue

From (attached) DAN clicking the button does not generate the decision tree.

Comments (3)

  1. Jorge Pérez

    The bug seems to be on DANDecisionTreeEvaluation method of the org.openmarkov.inference.decompositionIntoSymmetricDANs.DANDecisionTreeEvaluation class. In the line #67, the returned value (svNode) is null.

    DecisionTreeBuilder.getSuperValueNode(dan) when the dan only have one node (and it's potential) [see image] is returning that null.Screenshot_1.png

  2. Manuel Luque

    I've opened the network in text mode and I've found some problem in it.

    While the declared format version of the network is 0.2.0, the utility potential of node "Health state" appears as:

    <Potential type="Table" role="conditionalProbability">
            <UtilityVariable name="Health state" />
            <Variables>
              <Variable name="Therapy" />
              <Variable name="Disease" />
              <Variable name="Test A" />
            </Variables>
            <Values>9.3 9.9 10.0 4.0 6.5 1.2 9.299 9.899 9.999 3.999 6.499 1.199</Values>
          </Potential>
    

    I think in ProbModel 0.2 utility potentials had role "utility". However, in this network we can see that the role of the potential of "Health state" appears as "conditionalProbability".

    Inference tests are using the class PGMXReader_0_2 for loading tests networks. A ParserException is thrown when tests try to load this network.

  3. Log in to comment