基于定制时间约束的正负时态关联规则挖掘  

Customized Interval based Method for Mining Positive and Negative Temporal Association Rules

作　　者：董祥军[1,2] 陈建斌[1] 宋丽哲[1] 宋瀚涛[1] 陆玉昌[3] 

机构地区：[1]北京理工大学信息科学技术学院计算机工程系 [2]山东轻工业学院计算机科学与技术系,济南250100 [3]清华大学智能系统与技术国家重点实验室

出　　处：《计算机工程与应用》2004年第28期40-43,共4页Computer Engineering and Applications

基　　金：国家973重点基础研究发展项目(编号:G1998030414)

摘　　要：传统关联规则挖掘是在整个事务数据库的时间范围内进行的,但有时用户想得到某一特定时间范围(如商品的促销阶段)内的关联规则,该文对这一问题进行了详细讨论,提出了基于定制时间的时态支持度、时态频繁项集、时态置信度、时态关联规则等概念,在传统Apriori算法的基础上提出了挖掘时态频繁项集的算法。另一方面,讨论了当同时考虑正、负关联规则出现的矛盾规则问题以及用相关性解决这一问题的方法,提出了挖掘正负时态关联规则的算法,实例说明了算法的执行过程及有效性。The association rules are discovered traditionally in the interval of the whole transaction database.The users,however,sometimes are interested in the rules in a special interval(the sales promotion period for example).This prob-lem is discussed in details.Some concepts,such as temporal support,temporal frequent itemsets,temporal confidence and temporal association rule,are proposed based on customized interval.The traditional algorithm Apriori is modified to dis-cover temporal frequent itemsets.On the other hand,self-conflicting rules maybe occur when studying positive and nega-tive association rules simultaneously.This problem is discussed and the corresponding solution is given by applying cor-relation in association rules.An algorithm is proposed to mine positive and negative temporal association rules.An exam-ple is also given to demonstrate the algorithms' efficiency.

关 键 词：定制时间 负关联规则 时态关联规则 相关性 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]