基于量化相关模式的多值关联规则挖掘算法

Quantitative association rules mining based on quantitative correlated pattern

为了解决多值关联规则挖掘中忽视罕见且有价值的非频繁模式的问题,提出了一种新的多值关联规则挖掘算法—QCoMine。该算法引入了量化相关模式的概念,通过考察多值属性间互信息熵和全置信度,找到具有强信息关系的属性集进而产生规则。实验结果表明,由于在属性层和区间层进行了剪枝,因此缩减了搜索空间,提高了算法的性能,且得到更高置信度、更有价值的规则。

To resolve the mining problem of the quantitative association, which ignore the rare but much valuable non-frequent patterns, a new algorithm of quantitative association rules, QCoMine, is proposed. The new algorithm is based on a novel notion of quantitive correlated pattern, the mutual information entropy of the attributes and all-confidence are studied here, the attributes sets with strong information relationship is found. The expriments show that due to the prune on the attribute-level and the interval-level, the research space decrease sharply, so the mining efficiency is improved greatly, and the acquired association rules are high confidence and more valuable ones.