时间序列数据摘要与索引机制

A Time Series Data Summary and Index Mechanism

高容量的时间序列可视化分析在金融、离散制造等众多行业中应用越来越普遍,但高容量时间序列可视化在处理可视化延时需求和高摄取率上比较困难。针对该问题,提出一种基于M4的概要信息聚合预存储和索引机制。首先,将原始的时间序列数据根据不同压缩比例分层处理,每层处理中数据等距离分割成若干段,并选择每段中时间最大最小和序列值最大最小时对应的4个最值点;其次,将聚合后的数据存储在时间序列概要树上;最后,实现基于概要树索引查询算法。通过模拟生成时序数据集进行了实验分析,结果显示该索引机制与原有的M4聚合查询方法可达到更好的查询效果。

This paper proposes a method that is the M4-based summary information pre-store and index mechanism.First,compressing the original time series data by different compression ratio hierarchical,dividing the data into equidistance segments in each layer,and select the maximum minimum sequence,the largest and most hours value in each period of time.Second,insert the aggregated data into a time sequence summary tree.Last,implementing the index query algorithm based on summary tree.