From Intent Reducts for Attribute Implications to Approximate Intent Reducts for Association Rules

Automatic extraction of data dependencies or regularities from historical data is of great importance in practice. Concept lattice could serve as a common framework for such tasks. To deal with two typical kinds of data dependencies, attribute implications and association rules, this paper presents the definitions of intent reducts and approximate intent reducts, as the theoretical foundation for extracting these two kinds of knowledge. Algorithms for calculating them are developed. We also given out the methods for extracting attribute implications and association rules. The results provide a better understanding of attribute implications, association rules, and concept lattices