CAREER: Mining structure and dynamics of groups of nodes in real-world networks

社会、技术、信息和生物系统可以作为图形来研究，其中节点代表实体（即人、网站），边缘代表交互（友谊、通信）。该项目旨在分析和发现网络系统的解释和预测模型，例如大型人群和社会，或大型生物和技术系统，以了解其结构并对其全球动态进行预测。研究节点群体的结构和动态，旨在发明新的网络群体检测方法，建立节点群体行为的预测模型。提出的研究主要有三个重点：(1)网络社区的结构和发现；(2)网络社区的动态和“健康”；(3)具有丰富节点和边缘元数据的网络中的监督社区检测。研究的重点是利用大量的网络数据集，因为某些行为和模式只有在数据量足够大时才能观察到。该项目的智力重点是增加模型的表达能力，以包括丰富的节点和边缘元数据，并探索网络结构与节点和边缘的属性/特征之间的联系。教育计划提供了丰富的研究经验，并帮助学生发展跨学科的态度和技能，需要通过课程，看看现实世界的网络问题和数据。该提案的一个组成部分是公开发布用于分析大型网络的数据集和计算工具。有关该项目的其他信息，包括出版物、数据集、源代码和教育材料，可通过该项目的网站http://snap.stanford.edu访问。

Social, technological, information and biological systems can be studied as graphs, where nodes represent entities (i.e., people, websites) and edges represent interactions (friendships, communication). The project aims to analyze and discover explanatory and predictive models of networked systems, such as large groups of people and societies, or large biological and technological systems, in order to understand their structure and make predictions about their global dynamics.The research studies the structure and dynamics of communities of nodes, with the goal to invent novel network community detection methods and build predictive models of behavior of groups of nodes. The proposed research has three main thrusts: (1) Structure and discovery of network communities, (2) Dynamics and "health" of network communities, and (3) Supervised community detection in networks with rich node and edge metadata. The research focuses on harnessing massive network datasets, as certain behaviors and patterns are observable only when the amount of data is large enough. The intellectual focus of the project is on increasing the expressivity of the models to also include rich node and edge metadata and explore the connections between the network structure and the attributes/features of nodes and edges.The education plan provides for rich research experiences and helps students develop the interdisciplinary attitudes and skills needed for this work through courses that look at real-world network problems and data. An integral part of this proposal is public release of datasets and computational tools for analysis of large networks. Additional information about the project including publications, data sets, source code, and educational materials can be accessed through the project website at http://snap.stanford.edu.