Efficient indirect association discovery using compact transaction databases
Qian Wan, Aijun An · 2006
Abstract — An indirect association is a special type of negative association that relates two items via a mediator. The two items in an indirect association are rarely present together, but each of them occurs frequently together with the mediator. In this paper, we propose HI-mine*, an innovative optimization of the previously developed HI-mine algorithm for fast extracting indirect associations. This optimization is based on a novel strategy for compressing a transaction database into a Super Compact Transaction Database, which dramatically reduces not only the number of transactions in the database, but also the memory requirement for storing frequent-item projections in mining indirect associations. Our experimental results show that the HI-mine * algorithm is effective and efficient, and improves the performance of indirect association mining significantly. Index Terms — Data mining, association rule, indirect associa-tion, algorithm. I.