大数据及其科学问题与方法的探讨

大数据及其科学问题与方法的探讨 The Scientific Problems and Methodology of Big Data 何非 HE Fei 1 first-author 何克清 HE Keqing 2 武汉大学国际软件学院 武汉大学国际软件学院 International School of Software,Wuhan University International School of Software,Wuhan University 武汉大学软件工程国家重点实验室 武汉大学软件工程国家重点实验室 State Key Laboratory of Software Engineering,Wuhan University State Key Laboratory of Software Engineering,Wuhan University 大数据的规模和复杂度的增长超出了计算机软硬件能力增长的摩尔定律,对现有的IT架构以及计算能力带来了极大挑战,也为人们深度挖掘和充分利用大数据的大价值带来了巨大机遇.因此,针对网络大数据的共性规律和科学问题,需要研究网络大数据定性定量分析的基础理论与方法.本文重点分析了大数据在数据关联上的复杂性、计算复杂性、系统的复杂性、学习复杂性四个方面的主要研究问题和现状,介绍小结了产业界具有代表性的实际应用IT基础架构,以形成从网络大数据的感知、挖掘获取、质量评估、数据提炼处理到融合表示的综合过程图景,最后对大数据科学、数据计算需要的新模式与新范式（第四范式）等方面的发展趋势和方向进行了展望探讨. The increase of scale and complexity of Big Data exceeds that of the capacity of software and hardware characterized by the Moore law,which brings big challenge to the architecture,the processing and computing capacity of the contemporary IT systems,meanwhile presents unprecedented opportunities on deeply mining and taking full advantage of the value of Big Data.Therefore,it is pressing to investigate the disciplinary issues and discover the common laws of Big Data,and further study the fundamental theories and basic approaches to qualitatively or quantitatively dealing with Big Data.This paper analyzes the challenges caused by the complexity,uncertainty and emergence of Big Data,and summarizes major issues and research status of the data complexity,computing complexity,system complexity and learning complexity of Big Data.It also looks ahead to the development trends of Big Data science,as well as new modes,paradigm and IT infrastructure of data computing. 大数据 科学问题 科学方法 第四范式 本体元建模 big data scientific problems scientific methods the fourth paradigm ontology based metamodelling 10.14188/j.1671-8836.2014.01.001 TP311.13 国家重点基础研究发展计划973项目（2014CB340401）;国家自然科学基金资助项目（61373037） 2014-01-01 2021-04-01 1 60