Hello RapidMiner Team
I am currently using rm4.6. I am not sure whether this is a bug or not, but if so and you have already fixed it in 5.0 I will remain silent and go back to my cave
.
I am talking about the function: getInputDescription(Class cls), especially in the case of BinominalClassificationPerformance aka AbstractPerformanceEvaluator . If I did understand this method it checks whether an ExampleSet (in the mentioned special case) is in the "input-area" and if this is true, it creates the parameter "keep_example_set".
So ... if I create a new process and add BinominalClassifcationPerformance, no parameter is shown. Accepted. However, adding more operators ...
<operator name="Root" class="Process" expanded="yes">
<operator name="ExampleSource" class="ExampleSource">
<parameter key="attributes" value="/home/steffen/rm_workspace/sample/data/golf.aml"/>
</operator>
<operator name="NaiveBayes" class="NaiveBayes">
<parameter key="keep_example_set" value="true"/>
</operator>
<operator name="BinominalClassificationPerformance" class="BinominalClassificationPerformance">
<parameter key="AUC" value="true"/>
</operator>
<operator name="Dummy_to_show_error_message" class="BinDiscretization">
</operator>
</operator>
does not change what I see. Is this a bug ?
Before I got problems with the usage of XVPrediction. In detail, the following situations occur, but I havent see a pattern yet (yes, I changed the process, no, as far as I remember, I did not change the critical part):
- keep_example_set is shown
- keep_example_set is not shown and the process crashes because the exampleset is missing
- keep_example_set is not shown, the validation-check is d'accord and the process run smoothly
Adding IOMultiplier of cause solves the problem.
kind regards,
Steffen
PS: If I find a pattern, I will report it. But maybe this issue is already fixed in 5.0.
