Heinrich-Heine-Universität Düsseldorf

This paper describes a novel market basket dataset generator, which models various temporal and general aspects from real life. Besides taking into account general customer behavior, customer group behavior and several other aspects, the created datsets are timestamped in a manner that allows temporal data mining with regard to three major kinds of temporal association rules, namely cyclic, calendar- and lifespan-based association rules. The appearance of all these aspects can be controlled by the use of accordant parameters, which brings the user in total control of every little facet of the emerging datasets. This data provides an excellent basis for developing and testing market basket analysis algorithms.