Discovering Association Rules Change from Large Databases

Abstract

Discovering association rules and association rules change (ARC) from existing large databases is an important problem. This paper presents an approach based on multi-hash chain structures to mine association rules change from large database with shorter transactions. In most existing algorithms of association rules change, the mining procedure is divided into two phases, first, association rules sets are discovered using existing algorithm for mining association rules, and then the association rules sets are mined to obtain the association rules change. Those algorithms do not deal with the integration effect to mine association rules and association rules change. In addition, most existing algorithms relate only to the single trend of association rules change. This paper presents an approach which mines both association rules and association rules change and can mine the various trends of association rules change from a multi-hash chain structure. The approach needs only to scan the database twice in the whole mining procedure, so it has lower I/O spending. Experiment results show that the approach is effective to mine association rules using the multi-hash chain structure. The approach has advantages over the Fp-growth and Apriori algorithm in mining frequent pattern or association rules from large databases with shorter transaction. Besides, the experiment results also show that the approach is effective for mining association rules change and it has better flexibility. The application study indicates the approach can mine and obtain the practicable association rules change.