基于覆盖度函数的概念格约简coveragedensity-电子与信息学报.pdfVIP

第32 卷第2 期 电 子 与 信 息 学 报 Vol.32No.2 2010 年2 月 Journal of Electronics Information Technology Feb. 2010 基于覆盖度函数的概念格约简 姜 峰 范玉顺 (清华大学自动化系国家CIMS 工程技术研究中心 北京 100084 ) 摘 要：该文针对基于概念格的大规模数据和规则挖掘中，概念节点数呈指数爆炸的问题，提出了概念覆盖度函数 和概念格度量模型，进行概念格约简，从而使生成的标示概念格具有线性空间复杂度。给出了概念格约简的直求法、 同步法和提取法3 种算法。时空复杂度分析和仿真试验表明，所提方法可以大幅约简概念格规模，从而显著提高建 格和规则挖掘效率。标示概念还具有特殊含义，在Web 服务关系挖掘中有很好的应用。 关键词：数据挖掘；形式概念分析；概念格；约简；Web 关系挖掘 中图分类号：TP391; TP274 文献标识码：A 文章编号：1009-5896(2010)02-0405-06 DOI: 10.3724/SP.J.1146.2009.00099 Coverage Density Based Approach for Concept Lattice Reduction Jiang Feng Fan Yu-shun (Department of Automation, National Engineering Research Center for CIMS, Tsinghua University, Beijing 100084, China) Abstract: To address the lattice size exponential explosion problem in large scale data and rule mining, concept coverage density function and measurement model are introduced to reduce redundant concepts. The pruned lattice, named marked-concept lattice, has linear space complexity and can be obtained through direct or synchronous construction or node-extraction. Analysis and simulation tests show that this reduction model not only significantly reduces normal concept lattice size, but also significantly improves lattice building and rule mining efficiency. Furthermore, marked concept carries crucial information and physical meanings, thus can make benefits for Web service relationship mining. Key words: Data mining; Formal Concept Analysis (FCA); Concept lattice; Reduction; Web relationship mining [10] 1 引言 的效率。概念合并