科研之友
微信
新浪微博
Facebook
分享链接
摘要
Social network analysis, as an important part of complex network theories, attracts the attention of a large number of researchers. Different from the study of traditional social networks constructed by nodes of individuals and edges of social relationships, the study of modern social networks concerns not only static social relationships, but also dynamical interactions between people. Nowadays, the development of mobile internet and increasing popularity of smart mobile devices make it easier to sense human behaviors and interactions, whose temporal dynamics are mapped into dynamical social networks with variable topologies. In our study, we mainly focus on the temporal social networks constructed by human offline interactions, which require the geographical closeness between pairwise individuals in the physical world. As an important part of human dynamics and social computing, human offline interactions in dynamical social networks play a crucial role in lots of applications such as recommendation systems, epidemical immune strategies and opportunistic routings. In this paper, we review the offline interactions in dynamical social networks through four main facets: empirical data, characteristics, models and applications. Empirical data is essential to understand and analyze human interactions. Therefore, we discuss the features of different data sources and data acquisition methods such as WiFi, Bluetooth and GPS. In these data acquisition methods, both the latitude and longitude information identified by GPS, the wireless hotspot physical address in the WiFi connection log, the POS machines in the stores as the geographical tags are the proxy of human interactions. We classify interaction characteristics into three categories: topological, temporal and spatial, whose characteristics such as temporal clustering coefficients, temporal motifs and inter-contact time are described in detail. For human encounter modeling, we review three types of models based on traces, locations and topology, and introduce the state-of-art models in each type. Modeling offline interactions is a key step for simulating applications of offline interactions by making supplements of real world datasets. The artificial models, on the one hand, preserve the basic characteristics of real data, allowing the various parameters (e.g., the number of interactions, the number of locations, the length of time, etc.) to adjust in a more flexible way in the simulation situations. On the other hand, modeling offline human interactions helps to explore their essential mechanisms to provide theoretical support for the control and prediction in practical processes. With the abundance of data and the deepening of research, the study of dynamical social networks with offline interactions has brought a new perspective to the application of many fields such as inferring social ties, community detection, public health strategies, and recommendation systems. We finally outlook several directions with some open questions for future study.
- 出版日期2018
- 单位复旦大学
