基于相关兴趣度的关联规则挖掘算法研究

摘要

针对传统的基于支持度、置信度框架的关联规则挖掘技术的缺点和不足,文中在关联且相关的关联规则挖掘算法(association&correlation Mining,AC_Mining)的基础上,引入相关兴趣度——Related-confidence用于度量项集中项和项之间的相关性,提出一种新的挖掘算法I&ItemMine_AC(I&Item:项集和项,AC:association&correlation).实验证明,该算法消除了在传统关联规则挖掘中存在的可疑模式或关联规则,改善了一般性关联规则在挖掘前、后项集不对称情况时的不足,提高了所生成关联规则的质量,且其相关度量具有很好的剪枝效果.%Aimed at the demerits of association rule mining technology based on the degree of support and confidence,and based on both association and correlation rule algorithm association & correlation mining (AC_Mining),this paper introduces a related-confidence to measure the correlation among items,and proposes a new mining algorithm,called the I&ItemMine_AC (I&Item:item set and item;AC:association & correlation).Experiments prove that the mining algorithm eliminates suspicious patterns or association rules existed in the traditional association rule mining.Meanwhile,it also improves the demerits of general association rules and the unbalance before and after item sets mine,thus,improving the quality of generated association rules,and its relevant measure is very effective in pruning.