Social Network Analysis and Mining

信息网络是现代信息体系结构的重要组成部分，代表着包括社会网络、生物网络和计算机网络在内的多种网络。对Facebook或Twitter等社交网络的分析已经引起了不同学科的研究人员的极大兴趣，并在许多领域取得了重要成果，如特定群体识别、信息传播、影响力计算和群体不稳定。 本研究的目标是分析和挖掘异质信息网络，即具有多于一种类型的节点和/或链路并且能够表示多模式数据的网络。研究的目标有三个：(1)处理链接预测和社区发现等异质信息网络分析中的一些关键问题；(2)研究挖掘三元和更一般的n元关联规则的问题，以便将研究结果进一步应用于异质信息网络分析和超立方体挖掘；(3)探索有助于更好地满足前两个目标的模式的附加操作。模式是一种知识制品，如使用数据挖掘和分析过程从数据中提取的关联规则或聚类。 这一研究项目的创新之处在于：(I)通过采用先进的数据挖掘和分析技术，包括形式概念分析，开发了一个分析具有复杂结构的社会网络的框架；(Ii)通过考虑异质信息网络的特点，为链接预测和社区发现问题提出了新的解决方案；以及(Iii)通过提出直接从多维数据计算三元和n元概念、生成器和关联规则的新方法，对形式概念分析理论做出了贡献。

Information networks are a key component of modern information architecture and represent a variety of networks, including social, biological and computer networks. The analysis of social networks like Facebook or Twitter has gained very large interests from researchers in different disciplines and lead to significant results in many fields such as specific group identification, information propagation, influence computation and group destabilization. The goal of the present research program is to analyze and mine heterogeneous information networks, i.e., the ones which have more than one type of nodes and/or links and could represent multi-modal data. The objectives are threefold: (1) handle some key issues in heterogeneous information network analysis such as link prediction and community detection, (2) study the problem of mining triadic and more generally n-ary association rules in order to further apply the findings to heterogeneous information network analysis and hypercube mining, and (3) explore additional operations on patterns that could be helpful to better meet the first two objectives. A pattern is a knowledge artifact like association rules or clusters extracted from data using data mining and analysis procedures. The novelty of this research program comes from (i) the development of a framework for analyzing social networks with a complex structure by adapting advanced data mining and analysis techniques, including formal concept analysis, (ii) the proposal of new solutions to the issues of link prediction and community detection by taking into account the peculiarities of heterogeneous information networks , and (iii) the contribution to the theory of formal concept analysis by proposing new procedures for computing triadic and n-ary concepts, generators and association rules directly from multidimensional data.