多度量空间数据索引技术研究

索引作为数据处理系统的重要组成部分，是加速大数据处理的重要手段之一。尽管现有的度量空间数据索引适用于任意数据类型，但一个索引结构只能索引单一数据类型。万物互联时代产生了类型多样且结构各异的海量数据，亟需多度量空间数据索引以统一管理多类型异构数据并挖掘其潜藏价值。本项目拟设计高效的多度量空间数据索引，以解决数据实时更新、规模庞大以及类型多变的挑战，并结合新表征、新智用、新环境和新硬件的相关技术继续深入研究基于图的多度量空间数据索引、AI赋能的多度量空间学习型索引、分布式的多度量空间数据索引以及GPU加速的多度量空间数据索引，力争在相应的理论与方法上取得突破。最后搭建一个集成各种所提出多度量空间数据索引的一体化数据智能索引系统，为今后的实际应用奠定坚实的基础。本项目的实施可以丰富数据索引技术研究成果，促进大数据查询与挖掘领域的进一步发展以及助力数字中国建设。

Indexing is an important part of data management systems, which can be utilized to accelerate big data processing. Although existing metric space data indexing is general and applicable to any type of data, one index structure is only built for a single data type. In the era of Internet of Everything, large volumes of data with different types and various structures are continuously being generated. Hence, multi-metric space data indexing is required to support the unified big data management and to effectively mine the hidden value of big data. This project aims to investigate indexing techniques on multi-metric space data in order to address the volume, the velocity and the variety challenges. In addition, by using the new representations, the new intelligent techniques, the new environments, and the latest hardwares, the project further explores graph based multi-metric space data indexing, AI enhanced multi-metric space data indexing, distributed multi-metric space data indexing, and GPU based multi-metric space data indexing. Finally, the project builds a unified intelligent data indexing system that integrates all kinds of the proposed indexes, which aims to offer a solid foundation for future practical applications. The success of the project can enrich the efforts on data indexing, accelerate the development of querying and mining big data, and contribute to the construction of Digital China.