Synopsis
Performs an example weighting based upon the residuals of an unweighted local polynomial regression.
Description
This operator provides a weighting scheme based upon correlation. It calculates the correlation of each attribute with the label attribute and returns the absolute or squared value as its weight.
Please keep in mind, that polynomial classes provide no information about their ordering, so that the weights are more or less random, because depending on the internal numerical representation of the classes. Binominal labels work because of the representation as 0 and 1, as do numerical ones.
Input
- example set: expects: ExampleSet
Output
- weights:
- example set:
Parameters
- normalize weights: Activates the normalization of all weights.
- squared correlation: Indicates if the squared correlation should be calculated.