Using generators for discovering certain and generalized decision rules

Certain and generalized decision rules are often used in rough sets applications. A number of algorithms have been proposed to discover these types of rules by means of Boolean reasoning. An alternative approach is based on data mining techniques and proved to be useful in the case of large data sets. In this paper, we apply data mining achievements in the area of concise representation of patterns to generation of all optimal certain and generalized decision rules. We prove that antecedents of such rules are generators. We use this result to offer adapted versions of the AprioriCertain and AprioriGeneralized algorithms that drastically reduce the number of useless candidate rules.