Towards Efficient Computing Consistent and Partially Consistent Extensions of Information Systems

Abstract

This paper provides new algorithms for computing consistent and partially consistent extensions of information systems. A maximal consistent extension of a given information system includes only objects corresponding to known attribute values which are consistent with all rules extracted from the original information system. A partially consistent extension of a given information system includes objects corresponding to known attribute values which are consistent to a certain degree with the knowledge represented by rules extracted from the original information system. This degree can be between 0 and 1, 0 for the full inconsistency and 1 for the full consistency. The algorithms presented here do not involve computing any rules true in a given information system. This property differentiates them from methods presented in the earlier papers which concerned extensions of information systems.