多源异构地学时序大数据分布式并行索引与相似性计算方法

分布式并行时序索引与相似性计算是大规模时序数据高效访问与挖掘分析的基础，但还存在索引摘要耦合度低、相似距离计算复杂、层状索引结构导致并发控制成本过高等问题，缺乏对多目标、多传感器、多分辨率、多类型的多源异构地学时序大数据的统一索引机制与相似性度量方法。项目拟采用可变间隔对象作为时序片段的索引摘要，建立基于可变间隔对象之间变换距离的相似性模型，实现对多源异构时序数据的相似性度量；将可变间隔对象的中值和角度投影到二维间隔点集平面，把时序相似性查询转变为该平面上简单空间范围查询，最后通过虚拟分解树将此空间范围查询转变为可并行的普通间隔关系算子集，突破层状结构对并行索引实现的限制，细化索引并行粒度的同时降低并发控制开销，提升分布式多核计算环境下时序数据索引与相似性计算性能，为分布式多核计算环境下的大规模多源异构时序大数据的快速访问与相似性计算提供可行、通用、高效的分布式并行解决方案。

Distributed parallel time series indexing and similarity computing are the cornerstone of efficient access and mining analysis for large-scale time series data. However, there are still some problems, such as lower coupling degrees of index summarization, huge amount of similarity distance computing, high concurrent control overheads caused by hierarchical index structure. The unified indexing mechanism and similarity measurement method for multi-sensor, multi-resolution, multi-type and multi-source heterogeneous geoscience time series data is still lacking. The project intends to use transformable interval object as the index summarization of time series fragment, construct the time series similarity model through the transformation distance between transformable interval objects, and realize the similarity measure of multi-source heterogeneous time series data. And then it projects the median and angle of a transformable interval object to a two-dimensional interval point set plane so that the time series similarity query is transformed into a simple spatial range query on the plane. Finally, the spatial range query is transformed into a set of parallel common interval relationship operators through the virtual decomposition tree. That way breaks through the limitation of the hierarchical structure to the parallel index implementation, refines the parallel granularity of the index, reduces the concurrent control overhead and further improves the performance of time series data index and similarity. It provides a feasible, universal and efficient distributed parallel solution for fast access and similarity computing of large-scale multi-source heterogeneous time series big data in distributed multi-core computing environment.