基于候选项目集特性的改进Apriori算法研究

Research on Candidate Items Based on Improved Apriori Algorithm

通过对Apriori和1-k-Apriori两种算法进行细致分析和深入研究,结合这两种算法的优点,提出了I1-k-Apriori算法.1-k-Apriori算法中利用Lk-1与L1相连接来得到候选项目集,但是,有些情况下,用这种方法生成的候选k项集数量过大,导致k项集的筛选代价太大.I1-k-Apriori算法根据k-1项集的特性和事务数据库中数据的特性来决定产生k项集的方法,可以有效避免由于Lk项数过多而影响运算效率.实验结果表明,I1-k-Apriori算法较大提高了运算效率.

The Apriori algorithm and the 1-k-Apriori algorithm are analyzed.I1-k-Apriori algorithm is proposed,in which the advantages of the above mentioned algorithms are used.The idea of 1-k-Aprior is to generate the Lk by combining Lk-1 and L1.However,the number of candidate Lk item in 1-k-Apriori may be far more than that in Apriori in some cases.I1-k-Apriori chooses the method of Lk generating according to the feature of the Lk-1 and the size of transaction database so that the over cost can be avoided.The resul...