(A and B) or (A and C) simplifies to A and (B or C), which is no DNF. However, applying Morgan's law this can be rewritten as ‘(‘A or (‘B and ‘C)) which is already generated (= term tuple 2 1) if class labels are reversed: if ‘(‘A or (‘B and ‘C)) then 0 else 1.
If accuracy of rule = x, the complementary rule is 1-x. This evaluation is quick.
Same principle applies to rule length 2 where (A and B) = ‘(‘A or ‘B) and other rule lengths e.g. term tuple 4 vs 1 1 1 1.
(A and B) or (A and C) simplifies to A and (B or C), which is no DNF. However, applying Morgan's law this can be rewritten as ‘(‘A or (‘B and ‘C)) which is already generated (= term tuple 2 1) if class labels are reversed: if ‘(‘A or (‘B and ‘C)) then 0 else 1.
If accuracy of rule = x, the complementary rule is 1-x. This evaluation is quick.
Same principle applies to rule length 2 where (A and B) = ‘(‘A or ‘B) and other rule lengths e.g. term tuple 4 vs 1 1 1 1.