Apply Numeric2Polynominal (or other preprocessing operators) on feature subsets

nicugeorgian

Hi,

I know that the operator Numeric2Polynominal maps all attributes' numeric values to the corresponding nominal values. If I understood correctly, it's only the attributes' values that are mapped, and not the label (whose numerical value is to be predicted).

Is there a way (or an operator) to map only the numeric values of some attributes to the corresponding nominal values? I would actually need it for all but one attribute.

Thanks,
Geo

IngoRM

Hi,

yes, for this purpose the operator "AttributeSubsetPreprocessing" can be used. The attributes on which the inner operators should be applied can be defined via regular expressions. This example applies a discretization on all attributes but on attribute "att3":

<operator name="Root" class="Process" expanded="yes">
    <operator name="ExampleSetGenerator" class="ExampleSetGenerator">
        <parameter key="target_function"	value="sum"/>
    </operator>
    <operator name="AttributeSubsetPreprocessing" class="AttributeSubsetPreprocessing" expanded="yes">
        <parameter key="attribute_name_regex"	value="att.*[^3]"/>
        <operator name="BinDiscretization" class="BinDiscretization">
        </operator>
    </operator>
</operator>

Cheers,
Ingo