Big Data Computing

388 Big Data Computing

Rule Induction

If-then rules are human readable representations of induced models. They may

sometimes be preferable to decision trees, which can result in very complex

and difcult trees. One way to obtain a more readable representation is rst

to learn a decision tree and second to translate the tree into a set of rules (e.g.,

C4.5 rules). Another alternative is to employ algorithms that directly learn sets

of rules. One approach for rule induction is the covering approach, which at

each stage identies a rule that “covers” some of the instances. The covering

approach considers each class and seeks a way of covering all instances in it, at

the same time excluding instances not in the class. Algorithms for rule induc- ...