Research on Frequent Itemsets Mining Algorithm based on Relational Database

Abstract

Mining association rules between items is an important research direction of data mining, and the relational database is the most popular database, so mining association rules in the relational database is a very important research direction. At present, neither the Apriori algorithm nor its improvements resolve some problems generating candidate itemset and scanning the transaction set repeatedly, which lead to low efficiency. This paper proposes the frequent itemsets mining algorithm based on relational database based on the study of those important mining association rules algorithms and the storage characteristics of the transaction set and items in the relational database, and presents its concrete implementation and its optimization method. This algorithm combines items in a transaction to generate itemsets and counts the same itemsets in all transactions, which improve the efficiency of execution. Moreover, this algorithm doesn’t produce candidate itemsets, and only scans transaction database once, so promotes considerably efficiency. The result of experiments shows that, the frequent itemsets mining algorithm based on relational database has higher efficiency than the classical Apriori algorithm under certain conditions.