面向时空聚集特征的大规模社交网络空间社区探测与识别方法

In recent years, large social networking services (SNSs) have made a rapid development and swept the world. Online society of individuals which are called virtual communities with complex structures and different functions were born. Among them, some communities tended to be aggregated in time and space because of the relationship in geography, work or blood, which have promoted the social interactions and the transformations of relationship between the physical space and virtual space, played an important role in the spread and evolution of hot issues between online and offline. Therefore, uncovering the temporal-space dependent communities in large-scale social networks is essential for the sustained, rapid and sound development of socioeconomics. This project firstly presents a method to uniformly organize the massive data of social interactions and human activities collected from large social networks. Based on the theories of behavioral geography and human dynamics, this project tries to mine the mobility patterns and the spatial-temporal characteristics of spatial communities such as the spatial-temporal constraints of interactions and the spatial-temporal similarities of activities. By jointly considering the topological, spatiotemporal, and semantic features, this project tries to present an approach for detection and identification of spatial communities from large social networks. This project there can support the understanding of the mapping relationships and interaction mechanism between the virtual space and physical space. In addition, this project will contribute to friend recommendation, targeted advertising, and grasp the spread discipline of hot issues in virtual space.

大型在线社交网络在网络空间诞生了结构复杂、功能各异的网络社区。其中，基于地缘、业缘、血缘，在时间和空间上具有聚集特征的网络社区，推动了虚拟网络空间与现实物理空间的紧密互动，在线上与线下社会关系的转化、网络热点事件传播演化为线下群体性事件的过程中扮演着重要角色。因此，从大型复杂社交网络中探测与识别此类社区对社会经济快速与健康发展至关重要。本项目以大型社交网络的海量时空信息作为切入点，对社交网络关系与活动数据进行统一组织和管理；基于行为地理学与人类动力学相关理论，挖掘群体行为模式，凝练空间社区时空特征(时空关系约束机制与时空活动相似性)；提出融合网络拓扑结构、时空特征以及语义特征的空间社区探测方法，对社区属性和社区时空活动分布进行识别。本项目可为揭示网络空间和现实空间的相互映射关系与相互作用机理提供基础理论支撑；为社交网络好友推荐、精准广告投放、掌握热点事件时空传播路径等提供技术支持。

近年来，大型在线社交网络在网络空间诞生了结构复杂、功能各异的网络社区。其中在时间和空间上具有聚集特征的网络社区，推动了虚拟网络空间与现实物理空间的紧密互动，在线上与线下社会关系的转化、网络热点事件传播演化为线下群体性事件的过程中扮演着重要角色。本项目分别获取了大型社交网络腾讯QQ和新浪微博3000万用户的基本属性，位置信息，社会关系数据，深圳市上百万用户一个月的IC卡刷卡记录数据，采集了某社交媒体深圳市5000名用户一个月连续登录的位置数据，在多源城市时空数据分析框架的基础上，对社交网络关系与活动数据进行统一组织和管理，（1）发展社交网络多层次网络结构和高动态时空活动一体化的时空数据模型；（2）基于行为地理学与人类动力学相关理论，挖掘群体行为模式，提出基于变阶马尔科夫模型，利用群体轨迹辅助个体轨迹位置预测的方法；（3）研究群体出行活动与土地利用之间的相互关系；（4）提出融合网络拓扑结构、时空特征以及语义特征的空间社区探测方法；（5）发展多源数据融合方法。