Discovery of Abstract Knowledge from Non-Atomic Attribute Values in Fuzzy Relational Databases

In this paper we introduce attribute-oriented induction with partial vote propagation – a new approach allowing acquisition of generalized knowledge from uncertain data. We utilize a proximity-based fuzzy relational database as the medium carrying the original information, where the lack of precise information about an entity is reflected via insertion of multiple attribute values, and the fuzzy relation of α-proximity replaces the classical equivalence relation. Following a well-known approach for generalization of exact data in ordinary databases [9], we introduce a new method for induction of tuples with non-atomic attribute values. In our approach we take advantage of the implicit information about the generalized attributes in the fuzzy database model and apply it to generalize imprecise information.