missing values

Hi Udo,

there are some learners which directly supports missing values (but there is no switch for that). However, there is also a more generic approach (without the need for replacing the values by something like the average): you could just define a new "category" (let's call it "missing") and use the AttributeValueMapper to map all "?" to this new (artificial) category (this of course only works on nominal data). Then you can apply all learners and they will take the "missing" information into account.

In the following example, I replaced the nominal missing values with the new category "missing" and the numerical missings simply with the average. As you can see, the decision tree learner actually uses the new category (and it is also really important to use this information as you can easily see):

<operator name="Root" class="Process" expanded="yes">
    <operator name="ExampleSource" class="ExampleSource">
        <parameter key="attributes"	value="../data/labor-negotiations.aml"/>
    </operator>
    <operator name="NominalFeatureIterator" class="FeatureIterator" expanded="yes">
        <parameter key="type_filter"	value="nominal"/>
        <operator name="NominalMissingCategory" class="AttributeValueMapper">
            <parameter key="attributes"	value="%{loop_feature}"/>
            <parameter key="replace_by"	value="missing"/>
            <parameter key="replace_what"	value="?"/>
        </operator>
    </operator>
    <operator name="NumericalFeatureIterator" class="FeatureIterator" expanded="yes">
        <operator name="MissingValueReplenishment" class="MissingValueReplenishment">
            <list key="columns">
              <parameter key="%{loop_feature}"	value="average"/>
            </list>
        </operator>
    </operator>
    <operator name="DecisionTree" class="DecisionTree">
    </operator>
</operator>

Hope that helps,
Ingo