基于结构的网络数据统计分析

With implementation of the national “Internet Plus” action plan and the big data strategy, the “Internet of Everything” (IOE) is here and it is growing rapidly. People, goods and things are linked together, forming a large-scale and complicated social network, where massive high-dimensional network data are attached to. Different from traditional data, network data have two new features, saying complicated structure and sparse value, which gradually neutralizes the usage of classical theory of statistics, especially while handling the complicated correlation structure. To improve this situation, the demand for developing a series of new statistical methods and theories that are based on network structure is becoming more and more intense, which should receive sufficient attention in the academic circles. However, the current researches on network data and the corresponding complicated correlation structure are still in their infancy, which are a far cry from meeting the ever-growing actual demand. On this ground, guided by “Internet Plus” action plan and with the general theory of statistics at its core, this project will start with exploring the generative mechanisms of a variety of network structures, and then create a series of reasonable statistical models for the network data, followed by developing a series of new methods and theories for the statistical analyses of the network data. Its final purpose is to deeply understand, clearly describe and make the best of the complicated structure of network data, based on which a new pattern of statistics driven by network structure can be built and a lot of practical problems about the government administration and enterprise operation can be figured out as well.

随着国家“互联网+”行动计划和大数据战略的实施，互联网正快步走向“万物互联”，人、物、事都被链接起来，形成了规模庞大、结构复杂的社会网络，其中附着了海量、高维的相关数据。与传统数据不同，网络数据具有结构复杂、价值稀疏等新特征，由此经典统计学理论逐渐失效，尤其在处理复杂结构时遭遇了瓶颈。为突破这一瓶颈，发展基于网络结构的统计学新方法、新理论成为迫切需要解决的关键问题。然而，现存的关于网络数据及其复杂结构的研究尚处于起步阶段，还远远不能满足日益庞大的实际需求。鉴于此，本项目以“互联网+”战略为指引，以统计学原理为核心，从探索网络结构的生成机制出发，旨在建立科学合理的网络数据模型、提出有效的网络数据统计分析方法与理论；最终目的是深刻理解、明确刻画、充分利用网络数据的复杂结构，借以开创由网络结构为驱动的统计学新格局，同时致力于解决政府管理和企业经营所面临的一些实际问题。

本项目服务国家“互联网+”行动计划和大数据战略，重点研究“万物互联”背景下不断涌现出的规模庞大、结构复杂的各种网络数据。与传统数据不同，网络数据具有结构复杂、价值稀疏等新特征，由此经典统计学理论逐渐失效，尤其在处理复杂结构时遭遇了瓶颈。为突破这一瓶颈，本项目提出了一系列基于结构的网络数据统计分析方法和理论，主要包括网络结构探索与结构降维、基于网络结构的统计分析、结构相关的数据分析与应用三个方面，以统计学原理为核心，从探索网络结构的生成机制出发，建立了一些科学合理的网络数据模型、提出了有效的网络数据统计分析方法与理论，通过深刻理解、明确刻画、充分利用网络数据的复杂结构，发展了由网络结构为驱动的统计学研究框架，同时还利用相关研究成果解决了政府管理和企业经营所面临的一些实际问题。