A new algorithm for finding association rules

摘要

Finding association rules is mainly based on generating larger and larger frequent set candidates, starting from frequent attributes in the database. The frequent sets can be organized as a part of a lattice of concepts according to the Formal Concept Analysis approach. Statistics on the database contents can estimate the number of elements of the maximal frequent set as well as the frequent attributes. Since the lattice construction is database contents-dependent, the pseudo-intents (see Formal Concept Analysis) are avoided. They can be taken under consideration in the association rule generation step. Comparisons with the classical algorithms will be taken into account. The application applies to Oracle databases, via SQL queries.