信息网络关联关系分析技术研究

Information networks are ubiquitous in many applications and association relationship analysis on such networks has attracted significant attention in the academic communities. In recent years, various approaches have been proposed to deal with a variety of association relationship related research problems, including relationship identification, relationship closeness measuring，relationship prediction，abnormal or local aggregated relationship analysis. However, the existing technologies can be infeasible and inefficient when, as in many real-world scenarios, the networks is heterogeneous, dynamical, and in large-scale. This project will propose new approaches, or adapt existing technologies to fit the characteristics of complex network environments. Some key technologies such as dominate set based relationship summary structure generation, non-iterative relationship closeness computation and update strategies, composite relationship definition, general linear model based relationship prediction, mixed probability model based multi-relational collaborative clustering, significantly changed connected sub-graph extraction, and relative index based sub-graph similarity search will be explored. By this way, we can solve the problems of association relationship representation, measuring, prediction, abnormal and local aggregation analysis, and to promote and facilitate the development of association relationship analysis techniques over information networks.

信息网络关联关系分析技术是国际上新兴的研究方向，其在实践中具有非常广泛的应用。目前信息网络关联关系分析技术在关系识别、关系抽取、静态的关系紧密度度量等方面，已经取得了许多有价值的研究成果。然而在面向新型网络的复杂关系分析方面仍然存在着诸多不足。网络的大规模、动态及异构特性，使得关联关系的表示、度量、预测、局部聚集及异常分析面临巨大挑战。因此，本项目拟深入研究大规模动态异构环境下的信息网络关联关系分析技术，拟提出基于控制集的关系概要生成方法、关系紧密度的非迭代计算方法和动态维护策略、综合不同类型关系影响的复合关系定义方法、基于通用线性模型的关系预测技术、面向混合概率模型的多类型关系协同聚簇技术、变化显著的连接子图的快速抽取机制、基于相对索引的子图相似性搜索方法等，解决关系的表示、度量、预测、异常及局部聚集性分析等问题，促进和推动信息网络关联关系分析技术的发展。

本项目围绕信息网络关联关系分析技术展开了全面和系统的研究工作,并按照计划完成了全部研究工作。在网络的关系拓扑概要表示和压缩、关系紧密度的计算和动态更新、关系的预测及异常分析、关系的局部聚集性分析这4个方面取得进展。共有33篇学术论文被发表和录用，5篇论文被SCI收录。其中3篇发表在数据库领域重要国际会议ICDE和DASFAA上。2篇发表在著名国际期刊《Information Science》上。7篇发表在知名国际会议WAIM、APWEB、ADMA上。5篇发表在Frontiers of Computer Science、软件学报、计算机研究与发展等国内著名期刊上。其中论文“Probabilistic SimRank computation over uncertain graphs”是一篇扩展版ESI高被引论文（扩展至前3%）。论文“Accuracy Estimation of Link-Based Similarity Measures and Its Application”获得第十五届网络时代信息管理国际会议WAIM2014优秀论文亚军奖。结合理论研究成果，研制了国内第一个具有自主版权的集成化的个性化新闻推荐系统CROWN。描述该原型系统的demo paper "CROWN: Context-Aware Recommender for Web News" 已发表在ICDE2015国际会议上。结合本项目申请了3项专利，申请到了1项973计划子课题和1项863计划子课题。培养了4名博士生和18名硕士生。