An improved Apriori algorithm for mining association rules

Abstract

Among mining algorithms based on association rules, Apriori technique, mining frequent itermsets and interesting associations in transaction database, is not only the first used association rule mining technique but also the most popular one. After studying, it is found out that the traditional Apriori algorithms have two major bottlenecks: scanning the database frequently; generating a large number of candidate sets. Based on the inherent defects of Apriori algorithm, some related improvements are carried out: 1) using new database mapping way to avoid scanning the database repeatedly; 2) further pruning frequent itemsets and candidate itemsets in order to improve joining efficiency; 3) using overlap strategy to count support to achieve high efficiency. Under the same conditions, the results illustrate that the proposed improved Apriori algorithm improves the operating efficiency compared with other improved algorithms.