The final method will use a variant of the weight of evidence measure,
described in 
, to vote. The method there
cannot be used directly, since the weight of evidence must be
calculated or given in the rules for each attribute value describing a
class. The rules available from RGEN have accuracies and numbers for
the combination of several attributes, and not for each attribute
alone. We will instead use the J-measure from Equation as
shown in Figure .
Figure: Pseudo-code for an algorithm which selects a classification based
on weight of evidence.
will be equal to the accuracy of the current rule. will be given in the rule file as well, and is common for all rules with the same decision.