一种多最小支持度加权关联规则挖掘算法

针对实际交易数据库中,不同项目的重要性和出现概率各不相同的两个问题,提出一种基于等价类和多最小支持度的加权关联规则算法,从而挖掘出那些覆盖较少数据但却有意义、用户可能更感兴趣的关联规则。算法按照项目的最小支持度升序对交易记录进行等价类划分,然后按照项目的最小支持度降序依次求出每一等价类内的加权频繁项集。算法采用垂直数据库的数据表示形式,挖掘过程中避免了对数据库的重复扫描。对比实验结果证明,改进算法具有良好的挖掘性能。

Algorithm of mining weighted association rules with multiple minimum supports

Algorithm of mining weighted association rules based on equivalence classes and multiple minimum supports is proposed to solve the problem that the transactions and data items have not the same importance and frequency in datasets. The algorithm allows the user to specify varied minimum supports and gives items weights to find association rules those cover less data but are useful, in which the user are more interested. The algorithm divides the items into different equivalence classes in the ascending order of their MIS (minimum item supports) values, and then mines the weighted frequent itemsets in each class respectively in the descending order of the minimum support of items. Moreover, the algorithm adopts vertical database to represent primitive transactional database and there is no need to scan the database repeatedly. The experimental results show that the proposed algorithm is efficient.