模糊关联规则结合动态树重建的数据流挖掘  

Data Stream Mining Using Fuzzy Association Rules and Dynamic tree Reconstruction

作　　者：魏怀明 WEI Huai-ming(Department of Computer Engineering,Shanxi Architectural College,Taiyuan,Shanxi,030006,China;School of Computer Science and Technology,Tianjin University,Tianjin,050100,China)

机构地区：[1]山西建筑职业技术学院计算机工程系 [2]天津大学计算机科学与技术学院

出　　处：《控制工程》2018年第12期2263-2268,共6页Control Engineering of China

基　　金：国家自然科学基金面上项目(No.61373035)

摘　　要：针对基于频繁模式树的规则挖掘方法计算负载高和难以处理数据流格式,提出一种利用动态树构建的模糊关联规则挖掘方法。该技术整合了无处不在数据挖掘(UDM)和模糊集概念,首先利用滑动窗口最小化模糊关联规则。然后推导变量的模糊集,并给予适当描述,同时估计隶属度函数。最后构建动态树,并给每个节点添加一种隶属度函数值。根据当前窗口进行模糊关联规则推理。实验利用两种不同的公开数据,“交通事故”和“零售”数据集。考虑了3个模糊区域和五个模糊区域的运行分布。与频繁模式树(FPT)、压缩模糊频繁模式树(CFFPT)和熵加权频繁模式树(EWFPT)相比,提出的方法检索数据库只需要一次,且处理的数据要求更为宽松。Concerning that calculation of the rule mining method based on the frequent pattern tree is high and the method is difficult to handle the data stream format, the fuzzy association rule mining method constructed by a dynamic tree is proposed. The technology integrates ubiquitous data mining(UDM) and the concept of fuzzy sets. Firstly, the sliding window is adopted to minimize fuzzy association rules. Then the fuzzy set of variables is reasoned, and the appropriate description is given, and the membership function is estimated. Finally, a dynamic tree is built, and each node is added to a membership function value. The fuzzy association rule is reasoned based on the current window. Two kinds of public data sets "traffic accident" and "retail" are used in the experiment. The operation distribution of three fuzzy regions and five fuzzy regions is considered. Compared with frequent pattern tree(FPT), compressed fuzzy frequent pattern tree(CFFPT) and entropy-based weighted frequent pattern tree(EWFPT), the proposed method only needs one time, and the data request is more relaxed.

关 键 词：关联规则挖掘 频繁模式树 隶属度函数 动态树 数据流 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]