A prototype genetic algorithm-enhanced rough set-based rule induction system

The ability to team from empirical data or the observation of a real world and accurately predict future instances is an important feature expected in human. It allows knowledge to be gained by experience and decision rules i nduced from empirical data. One of the major obstacles in performing rule i nduction from empirical data is the inconsistency of information about a pr oblem domain. Rough set theory provides a novel way of dealing with vaguene ss and uncertainty. When coupled with genetic algorithms, a rule induction engine that is able to induce probable rules from inconsistent information can possibly be developed. This paper presents an integrated approach that combines rough set theory, genetic algorithms and Boolean algebra, for indu ctive learning. Using such an approach, a prototype system (RClass-Plus) th at discovers rules from inconsistent empirical data, has been developed. Th e system was validated using the data obtained from a case study. The resul ts of the validation are presented. (C) 2001 Elsevier Science B.V. All righ ts reserved.