时空异常值检测的研究

摘要

异常值检测是数据挖掘研究领域的一个相当重要的分支，异常值检测的目的是寻找与其他大多数对象不同的个体，又常被称为离群检测或者是例外挖掘。许多文献已经对空间异常值检测和时间序列异常值的检测进行了研究，然而同时对时间维度和空间维度进行异常值侦测的还不多；本文将分别从时间和空间这两个维度出发对异常值的检测进行研究，然后再将这两个维度结合起来，提出一种全新的时空异常值的检测方法，为未来时空异常值的检测奠定基础。%Outlier detection is a very important branch of data mining, the purpose of outliers detection is to find out the individuals that are different from most other objects, it is often called outlier detection or exception mining. Many literatures have studied the detection of spatial outliers and detection of temporal outliers, but there are not too many methods to connect temporal outliers and spatial outliers. This paper will research temporal outliers and spatial outliers individually, and then combine these two dimensions to propose a new detection method of spatiotemporal anomalies, which will lay a foundation for the future detection of Spatio-temporal anomalies.