摘要： 以“k倍标准差”准则为基础，提出一种专题属性双重偏离的时空异常检测方法，在每个要素的空间邻近域里采用“k倍标准差”准则探测各时刻的空间异常数据，在每个空间异常数据的时间邻近域中，再次使用该准则判断该要素是否为时序异常，并将所有空间和时间邻近域上均表现为异常的数据定义为时空异常。实验结果表明，该方法是有效可行的。

关键词: 时空数据挖掘, 时空异常探测, k倍标准差, 双重偏离

Abstract: Depending on “k standard deviation” rule, a new Spatio-Temporal Outlier Detection(STOD) method is proposed, named as a Dual non-spatio-temporal attributes Deviation-based STO Detection(DDSTOD). For each time instant, in Spatial Neighborhoods(SN) of each feature, the rule is employed to judge whether the spatial feature is a spatial outlier, and all the spatial outliers form a spatial outlier set. In the Temporal Neighborhoods(TN) of each spatial outlier, the rule is utilized to check temporal outliers, too. All features which are outliers both in their SN and TN composite Spatio-Temporal Outliers set(STOs). Experimental results show this method is effective and feasible.

Key words: spatio-temporal data mining, Spatio-Temporal Outlier Detection(STOD), k standard deviation, dual deviation

中图分类号: 

• N945