跨地区短时间序列的离散化方法

Discretization of Cross-Sectional Short Time Series

跨地区时间序列是由多个地区组成的多变量时间序列集合,每个地区有着相同的变量集。其时序长度通常比较短。跨地区时间序列在商业、经济、公共管理等领域广泛存在。离散化是时间序列关联规则挖掘的不可或缺的先行步骤。该文介绍了一个新颖的针对跨地区时间序列的离散化方法。相对没有考虑跨地区性质的方法,在跨地区短时序数据上,此方法可以更准确地提取出事件,完成离散化工作。

The cross-sectional time series data means a group of multivariate time series each of which has the same set of variables.Usually its length is short.It occurs frequently in business,economics,governance,and so on.Discretization is the indispensable antecedent step of time series association rule mining.This paper introduces a novel method to discretize cross-sectional time series.The method groups sections by statistical identical testing,and distill proper events on each group.On cross -sectional short time series,this method behaves better than separately discretization method does,which is confirmed by experimental results on synthetic data.