基于时间窗口聚类的时序数据索引压缩

子序列匹配是时间序列挖掘的经典课题,旨在发现大型数据集中的相似数据序列.很多文献关注固定时间段的序列的查询.但对于多种不同时间段的查询的问题仍然未解决好.基于时间段的查询含义是有时间窗口限制的查询.为了满足多时间段上的查询,简单地为每个时间段的子序列构建索引既耗时又耗存储空间.从目前的文献来看,已有的索引无法满足具有不...

A CLUSTERING-BASED INDEXING APPROACH FOR VARIABLE WINDOW QUERY OF TIME SERIES

Subsequence matching is a classic topic of time series mining.Many works aim to discover similar data series among large datasets.Lots of literatures pay attention to queries over fixed time period.However,variable time windows query remains an unsolved problem.Window-based query means querying with window limitation.However,simply building indexes for each time period will be both time-consuming and space-consuming.In the variable window query situation,different indexes should be built for different windows.So,state-of-the-art indexes could not satisfy a large number of queries with different window limitations.In this paper,we propose CBI,a clustering-based variable-window indexing approach that reduces both indexing time consumption and space cost for variable-window queries.The experiment results show that our index saves time and space cost,at the same time,performs high efficiency on exact query,approximate query,and range query.