基于模糊逻辑的多维时序软关联规则挖掘  

Soften Association Rules Mining of Multivariate Time Series Based on Fuzzy Logic

作　　者：王炳雪[1] 陈元忠[1] 

机构地区：[1]上海财经大学信息管理与工程学院,上海200433

出　　处：《计算机工程》2011年第10期35-37,共3页Computer Engineering

基　　金：上海财经大学"211"工程三期基金资助项目

摘　　要：针对时间序列片段归类时在边界状态上存在不确定性的问题,提出一种软化边界的方法。该方法对时间序列记录集属性序列进行滑窗以及正规化处理后用模糊聚类方法聚类,使样本个体不是简单地归于单个代表形态。通过样本点的隶属度计算关联规则的支持度和可信度,使这2个重要指标的计算更精确,并采用一种基于隶属度的J-measure测度对规则有效性进行排序。实际算例显示该算法能提高可信度和J-measure测度。Aiming at that the boundaries of time series exists uncertainty when clustering,this paper presents a method of soften boundaries.In order to form sub-series,the values of multivariate time series are put into recordset＇s attributes and windows of given width are slided through the attributes.The sub-series are normalized with a simple method and then clustered with fuzzy logic so as to obtain its delegates with soften boundaries.Rule＇s support and confidence are calculated with membership so as to make the two important measures more exact.Good rules are selected with J-measure which is based on membership.Empirical results show that the patterns and rules are meaningful and resultful.

关 键 词：数据挖掘 多维时间序列 软关联规则 模糊逻辑 模糊聚类 

分 类 号：TP311.12[自动化与计算机技术—计算机软件与理论]