频繁模式挖掘的约束算法  

A frequent pattern mining algorithm based on constraints

作　　者：孟彩霞[1] 

机构地区：[1]西安邮电学院计算机系,陕西西安710065

出　　处：《智能系统学报》2009年第2期142-147,共6页CAAI Transactions on Intelligent Systems

基　　金：陕西省自然科学基金资助项目(2004f283);西安市科技创新支撑-应用发展研究计划资助项目(YF07024)

摘　　要：在频繁模式挖掘过程中能够动态改变约束的算法比较少.提出了一种基于约束的频繁模式挖掘算法MCFP.MCFP首先按照约束的性质来建立频繁模式树,并且只需扫描一遍数据库,然后建立每个项的条件树,挖掘以该项为前缀的最大频繁模式,并用最大模式树来存储,最后根据最大模式来找出所有支持度明确的频繁模式.MCFP算法允许用户在挖掘频繁模式过程中动态地改变约束.实验表明,该算法与iCFP算法相比是很有效的.Most algorithms don＇ t allow users to dynamically change constraints in the process of mining frequent patterns. A new algorithm, constrain-based frequent patterns mining, was developed to provide frequent pattern mining with constraints. First, the algorithm constructs the FP-tree （frequent pattern tree） according to the descending or ascending order of constraints, and in this process the database only needs to be scanned once. Secondly, the conditional tree of each item was established to mine maximal frequent pattern with this term as a prefix, and the maximal frequent patterns were stored. Finally, all frequent patterns with precise support degrees were discovered according to the maximal frequent patterns. The significance of this method is that this algorithm allows users to dynamically change constraints during the process. Experimental outcomes showed that the proposed algorithm is more efficient than the algorithm of iCFP.

关 键 词：频繁模式挖掘 动态约束 频繁项集 最大频繁模式 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]