基于SFVS的时序关联规则动态发现方法  

Dynamically temporal association rules mining based on SFVS

作　　者：张新玉[1] 夏士雄[1] 牛强[1] 

机构地区：[1]中国矿业大学计算机学院,江苏徐州221116

出　　处：《计算机应用研究》2012年第7期2571-2574,共4页Application Research of Computers

基　　金：中国矿业大学青年科技基金资助项目(2011QNB23)

摘　　要：针对时间序列关联规则挖掘存在时间复杂度高、效率低等问题,将基于SFVS(统计特征矢量符号化)的时间序列表示方法引入到时序关联规则发现中,利用描述时序数据统计特征的均值与方差分别作为描述其平均值及发散程度的分量,实现时间序列表示的矢量化,然后再进行动态关联规则挖掘。实验结果表明,基于该方法所获取的关联规则具有更高的精确度和可信度。There are many disadvantages of discovering the association rules directly on the temporal series,such as high time complexity,low efficiency.So This paper considered that introduced the temporal series method based the SFVS（statistic feature vector symbolic） algorithm into the association rules discovering.Used the mean and variance describing the characteristics of temporal series data as the component describing the average and the degree of divergences,that was to make the temporal series be vectors.Then discovered the association rules dynamically.The experimental results show that the association rules discovered based SFVS algorithm have a higher accuracy and reliability.

关 键 词：时间序列 统计特征矢量 符号化表示 关联规则 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]