基于人机协同的海量数据探索系统关键技术研究

数据探索作为一种新兴的数据管理技术，被认为是提高大数据易用性的有效手段。针对海量数据的数据探索是在满足与海量数据交互性能的前提下，在用户在对数据域的认知不明确、对搜索目标信息及路径不确定的条件下，通过概要数据和人机协同的手段，有效地了解数据进而获取信息的过程。已有的数据探索系统不能完全把握以上的特征，所以在交互性能或者人机协同有效性上存在不足。本课题旨在研究基于人机协同的海量数据探索系统的若干关键技术；功能方面涵盖了数据探索系统在数据聚集、海量数据处理和人机交互这三大趋势；技术方面融合了MPP技术、在线聚集技术、数据质量管理技术和机器学习技术的优势；期望为解决大数据易用性问题提供新的理论体系和研究工具。

Data exploration is an emerging data management technique that enhances the usability of big data. In general, the data exploration technique refers to effectively understanding and accessing the target information through data synopsis and highly efficient human-computer interactions over massive datasets. More, such interactions are implemented while users are unfamiliar with data domains, unsure about the target information, and uncertain about searching paths. However, existing systems only partially address the above features and thus fall short in either interaction efficiency or human-computer collaboration effectiveness. Our research is focused on key technologies of a data exploration system with human-computer collaborations which covers the three major features of data exploration, including data aggregation, massive data management, and human-computer collaboration. The key technical part of our research deeply integrates the advantages of massive parallel processing, online aggregation, data quality management, and machine learning techniques. The achievements of our research will provide both theoretical and technical supports for improving the big data usability.