基于二维表的频繁集组合方法

挖掘关联规则时,频度集的计算是一个关键问题。现有算法大多从apriori,fp-growth算法演化而来,这些方法都存在有组合爆炸问题:apriori算法是候选模式造成的,而fp-growth算法是条件模式基造成的。因此,对于具有稠密数据的事务数据库现有的方法无法完成频度集的计算,即使是现有最好的fp-growth算法。将给出一种基于迭代思想的频度集计算方法,既不用频繁扫描数据又不用组合计算,实验表明此方法是十分有效的频度集计算方法。

Combining Frequent Sets Method on 2-Dimension Table

When we excavate connection rules,the frequency collection computation is a key problem.The existing algorithm evolves mostly from apriori and the fp-tree algorithms.These methods all involves combinatorial explosion:the apriori algorithm comes from the candidate pattern,but the fp-tree algorithm evolves from the condition pattern base.Therefore,regarding the routine database with dense data,the existing method is unable to complete the frequency collection computation,even if with the best existine fp-tree algorithm.This paper will produce a frequency collection computational method based on the iteration thought neither with the frequent scanning data nor with the combinatorial computation.The experiments show that this method is a very effective frequency collection computational method.