[Show abstract][Hide abstract] ABSTRACT: In this paper, the concept of a granulation order is proposed in an information system. The positive approximation of a set
under a granulation order is defined. Some properties of positive approximation are obtained. For a set of the universe in
an information system, its approximation accuracy is monotonously increasing under a granulation order. This means that a
proper family of granulations can be chosen for a target concept approximation according to the user requirements. An algorithm
based on positive approximation is designed for decision rule mining, and its application is illustrated by an example.
Rough Sets, Fuzzy Sets, Data Mining, and Granular Computing, 09/2005: pages 701-708;