基于效益度的高效关联规则挖掘算法

关联规则挖掘算法中常用的支持度和可信度是对关联规则在统计意义上的有效性度量，在挖掘结果的有用度上缺乏指导作用，它们不能作为有用性的指标．从数据挖掘的最终目的出发定义了基于最终用户实际目标的效益度指标，并对最小效益度筛选性质进行了论证，提出了一种快速有效的关联规则挖掘算法.讨论了从关联规则的兴趣模板和限制模板转换到效益度的方法.实验结果表明，效益度指标具有支持度与可信度不可替代的作用；该算法的最小效益度剪切技术是有效的，不仅可以较大幅度地提高算法速度，而且可以作为规则模板的统一实现算法以及提供更精确的控制.

Effective mining algorithm of association rules based on benefit measure

In association rule mining,the traditional statistical measures,such as support and confidence,tend to generate too many spurious patterns,which cannot reflect user's requirement accurately.In order to solve this problem and obtain rule knowledge based on end users' real demand,this work proposed a new benefit measure according to the view of decision-making.The anti-monotony property about the new measure was proved,and the relationship and transforming method between constraint template and benefit measure were investigated.Moreover,an effective association rule mining algorithm PCBARM was presented.Experimental results showed the effectiveness of the benefit measure and the efficiency of PCBARM algorithm.The proposed measure not only can be used as a general realization algorithm for rule template but also provides more precise control.