改进的多维关联规则算法研究及应用  

Research and Application of Improved Multidimensional Association Rule Mining Algorithms

作　　者：张素琪[1] 梁志刚[2] 胡利娟[2] 董永峰[2] 

机构地区：[1]天津大学,天津300072 [2]河北工业大学计算机科学与软件学院,天津300130

出　　处：《计算机工程与科学》2012年第9期174-179,共6页Computer Engineering & Science

基　　金：天津市自然科学基金资助项目(10JCZDJC16000)

摘　　要：关联规则是数据挖掘研究中最主要、最活跃的领域之一。以Apriori算法为前提,借助Apri-oriTid算法事务压缩的思想,减少了重复扫描数据库的时间;并提出了一种利用事务标识列表,该列表长度即是对应候选项集的支持度计数,在计算支持度计数时,仅需要得到对应列表长度即可,从而缩短了计算计数时的比较时间;同时,在生成频繁项集时引入地址索引机制,在剪枝过程中,利用候选项集的首元素在地址索引表中快速定位,减少了多次扫描事务数据库,有效地缩短了计数时间和占用的内存空间。利用改进的算法对科研管理系统数据进行关联关系分析,从中萃取数据中隐含的、有价值的信息,辅助下一阶段的科研管理工作。并通过试验进行性能比较得出,改进后的算法效率更高。The field of data mining association rules is one of the most important and active areas . Taking the Apriori algorithm as a premise , using the Affairs compression idea of AprioriTid algorithms, we reduce the duplication of time scanning the database. We put forward a kind of Apriori algo-rithm based on the identifier lists of transactions in the database, and the list length is the candidate sets＇ corresponding support count. For getting the support count in the calculation, we only need to count the length of the list, thereby reducing the calculation time. At the same time, introducing the address indexing mechanism when generating frequent itemsets in the pruning process, we use the first set of candidate elements in the address table index to quickly locate, and thus reduce the number of scanning the transaction database. We make use of the business address index table to improve the counting time and execution efficiency of algorithms. The data of scientific research management as the research object, we use the improved algorithms to analyze the data of relationship, moreover, to extract the data＇s hidden ,valuable information, and support the next phase of scientific research management. The experiments show that the algorithm is more efficient.

关 键 词：关联规则 数据挖掘 APRIORI算法 地址索引 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]