Granular Computing on Extensional Functional Dependencies for Information System

In this paper, a new approach to discover extensional functional dependencies for information systems is presented based on information granules using their bit representations. The principle of information granules, granular computing and the machine oriented model for data mining are investigated firstly. In addition, the approach to identify the classical functional dependencies, identity dependencies and partial dependencies is discussed and some conclusions on extensional functional dependencies are obtained. The information granules are represented with bit, then the data format can be closed to the inner representations of the computer, hence, the patterns contained in the information system can be directly mined.