Rule induction from inconsistent and incomplete data using rough sets

Proposes two methods based on rough sets theory to obtain minimal rules in an information system with inconsistencies and incompleteness. Both methods make use of the definition of a binary discernibility matrix to replace sets operations by bit-wise operations in the search of minimal coverings. The first method is an exhaustive search of coverings and the second uses a genetic algorithm (GA) based search. Inconsistencies are solved with the lower and upper approximations and the incompleteness problem is faced by modifying the definition of discernibility between pairs of examples into a rough discernibility (i.e. surely discernible and possibly indiscernible).