A framework of rough sets based rule generation in non-deterministic information systems

Rule generation in non-deterministic information systems, which follows rough sets based rule generation in deterministic information systems, is studied. For possible implications in non-deterministic information systems, six kinds of classes are newly introduced, and certain rules are defined by possible implications belonging to a class. Similarly, possible rules are defined by possible implications belonging to other three kinds of classes. Furthermore, the lower and upper approximations of criteria, i.e., support, accuracy and coverage of rules, are introduced. Finally a tool, which extracts minimal certain and minimal possible rules based on a total order over attributes, has been implemented.