Extracting association rules from a retail database: Two case studies

Abstract

An important issue related to database processing in retail organizations refers to the extraction of useful information to support management decisions. The task can be implemented by a particular group of data mining algorithms i.e., those that can identify and extract relevant information from retail databases. Usually it is expected that such algorithms deliver a set of conditional rules, referred to as association rules, each identifying a particular relationship between data items in the database. If the extracted set of rules is representative and sound, it can be successfully used for supporting administrative decisions or for making accurate predictions on new incoming data. This work describes the computational system S_MEMISP+AR, based on the MEMISP approach, and its use in two case studies, defined under different settings, related to the extraction of association rules in a real database from a retail company. Results are analyzed and a few conclusions drawn.