Possible equivalence relations and their application to hypothesis generation in non-deterministic information systems

Abstract

Non-deterministic Information Systems (NISs) are advanced extensions of Deterministic Information Systems (DISs), and NISs are known well as systems handling information incompleteness in tables. This paper examines manipulations on equivalence relations in DISs and manipulations on possible equivalence relations in NISs. This paper also follows rough sets based rule generation in DISs, and proposes rough sets based hypothesis generation in NISs. A hypothesis in NISs is defined by a formula satisfying some kinds of constraint, and effective algorithms for generating hypotheses in NISs are presented. Possible equivalence relations play important roles in generating hypotheses. Some illustrative examples and real executions of algorithms are shown, too. © Springer-Verlag 2004.