加权模糊关联挖掘算法_刘闻超.pdfVIP

( ) 计算机工程与设计 Computer Engineering and Design 3654 2010,31 16 计算机工程与设计 Computer Engineering and Design 智能技术 加权模糊关联挖掘算法 1 1,2 1 刘闻超 ， 施化吉 ， 马素琴 (1. 江苏大学计算机科学与通信工程学院，江苏镇江2 12013 ； 2. 南京航空航天大学计算机应用研究所，江苏 南京2 10016) 摘 要：针对数量型关联规则挖掘中划分边界过硬问题，以及加权关联规则中为确保向下封闭性成立而引起的规则丢失问 题，提出一种新的加权模糊关联挖掘模型及其挖掘算法NFWARM 。为了避免区间划分引起的边界过硬问题，该模型引入模 糊集软化属性的划分边界；同时，使用属性权重刻画元素对规则的贡献，在保证频繁项集向下封闭性的情况下，不会引起规 则丢失。实验结果表明，该算法适用于包含布尔型和数值型数据的大型数据库的规则挖掘，并且得到的频繁项目集数 目和 规则数目有显著增加。 关键词：数据挖掘; 加权关联规则; 模糊关联规则; 向下封闭性; 隶属度函数 中图法分类号：TP311 文献标识码：A 文章编号：1000-7024 (2010) 16-3654-04 Algorithm of weight fuzzy association rules 1 1,2 1 LIU Wen-chao , SHI Hua-j i , MA Su-qin (1. School of Computer Science and Telecommunication Engineering, Jiangsu University, Zhenj iang 2 12013, China; 2. Computer Application Institute, Nanj ing University of Aeronautics and Astronautics, Nanj ing 2 10016, China) Abstract ：To solve the problem of strong division in mining quantitative association rules and rules loss caused by ensuring the validation of downward closure property in weighted association rules, a new weighted fuzzy association rules mining model and its mining algo- rithm NFWARM are proposed. In the model, boundaries of division are softened using fuzzy set, and the problem of strong division is avoided; in the meantime, elements’contribution to rules is measured by attribute weight, and the problem of rules loss is avoided with the validation of downward