100 JOURNAL OF COMMUNICATIONS AND NETWORKS, VOL. 17, NO. 2, APRIL 2015 Impact of Human Mobility on Social Networks Dashun Wang and Chaoming Song Abstract: Mobile phone carriers face challenges from three syner- ical space—social connections between individuals and their gistic dimensions: Wireless, social, and mobile. Despite significant mobility—no longer exist in isolation. Rather they increasingly advances that have been made about social networks and human interact with and depend on each other. To truly harness and un- mobility, respectively, our knowledge about the interplay between leash the potential of social and mobile technologies, we need two layers remains largely limited, partly due to the difficulty in to develop a quantitative framework of the interplay between so- obtaining large-scale datasets that could offer at the same time so- cial networks and human mobility patterns. cial and mobile information across a substantial population over Our knowledge about the interplay between social networks an extended period of time. In this paper, we take advantage of a massive, longitudinal mobile phone dataset that consists of hu- and human mobility patterns is limited, partly due to the diffi- man mobility and social network information simultaneously, al- culty in obtaining large-scale dataset that could offer at the same lowing us to explore the impact of human mobility patterns on the time social and mobile information across a substantial popula- underlying social network. We find that human mobility plays an tion over an extended period of time. This situation is changing important role in shaping both local and global structural prop- drastically, however, thanks to the ever-increasing availability erties of social network. In contrast to the lack of scale in social of detailed traces of human behavior [13], [14], from mobile- networks and human movements, we discovered a characteristic phone records to global-positioning-system (GPS) data to loca- distance in physical space between 10 and 20 km that impacts both tion based social networking services. Take mobile phones as an local clustering and modular structure in social network. We also example: On one hand, mobile phones are carried by individu- find a surprising distinction in trajectory overlap that segments so- als during their daily routines, acting as an excellent proxy to cial ties into two categories. Our results are of fundamental rele- capture individual human trajectories; On the other hand, with vance to quantitative studies of human behavior, and could serve almost 100% mobile-phone penetration in industrial countries, as the basis of anchoring potential theoretical models of human be- mobile communications offer us a comprehensive picture of so- havior and building and developing new applications using social and mobile technologies. cial interactions within a society [10], [11], [15]–[24]. There- fore, these large-scale datasets, capturing time resolved loca- Index Terms: Clustering, heterogeneous network, human mobility, tions of individuals and their interactions, are not only fuel- mobile phones, percolation, scale-free network, social network. ing rapid advances in our understanding of individual mobility patterns, from empirical analyses to modeling tools and frame- works, but also offer unique opportunities to jointly explore in- I. INTRODUCTION dividual trajectories and social communications [8], [20], [25], [26]. Indeed, there has been growing evidence examining phys- OCIAL networks have attracted significant interest across ical space as a determinant of social ties. By assuming that in- Smultiple disciplines in recent years, largely due to their crit- dividuals have a fixed time-independent location in space, like ical role in a wide range of applications [1]–[7]. Despite recent their home or work addresses, previous research examined the explosion of research on social networks, the bulk of work has probability of a social tie between them given the distance be- primarily focused on the social space, leaving its interplay with tween their locations, finding a robust power law decay in such the physical space largely underexplored. Yet, in an accelerating probabilities across different contexts [27]–[30]. In the dynamic number of settings, we are witnessing emerging convergence in dimension of physical space, researcher discovered that, not social and mobile technologies, fueling rapid advances in areas only are individual trajectory overlaps predictive for the forma- as broad as marketing, security and communications. For exam- tion of social ties [9], [20], [31] location information of social ple, location-based social networking services offer information neighbors is also revealing about individuals’ whereabouts [8]. sharing that enables new ways in marketing, connecting with While these new results focused on dyadic, if microscopic, so- friends, and recommending services [8], [9]. Mobile phone car- cial interactions, they document first few connections between riers face challenges from three synergistic dimensions: wire- human mobility and social networks, portending to a quantita- less, social, and mobility [10]–[12]. Together these dimensions tive framework between these two areas. are imperative to determine their key service safety, operabil- Here, we take advantage of a massive, longitudinal mobile ity and profitability, from data transmission efficiency and ser- phone dataset and ask the question: to what degree do our mo- vice reliability to vulnerability and security. Therefore, as our bility patterns impact the dynamical and structural properties of society becomes globally interconnected, the social and phys- our social networks? On both areas of social networks and hu- Manuscript received January 9, 2015. man mobility, we establish a quantitative framework by carrying D. Wang is with the College of Information Sciences and Technology, out a series of analyses that attend to both static and dynamic Pennsylvania State University, University Park, PA 16802 USA, email: patterns. Through these analyses, we offer new evidence and in- [email protected]. sights of how social networks and human mobility correlate and C. Song is with the Department of Physics, University of Miami, Coral Gables, FL 3314 USA, email: [email protected]. interplay with each other in multiple levels of granularity. We Digital object identifier 10.1109/JCN.2015.000023 1229-2370/15/$10.00 c 2015 KICS WANG AND SONG: IMPACT OF HUMAN MOBILITY ON SOCIAL NETWORKS 101 Fig. 1. Three-month trajectory of one mobile phone users. find that human mobility plays an important role in shaping both mobile communicationwas initiated, the dataset also records the local and global structural properties of social network. We find informationof the mobile tower that routedthe phonecall or text the emergence of spatial segmentation that reveals higher order message, whose geographic coordinates allow us to pinpoint the regularities behind previously well-established results by study- location of the individual at the time when s/he initiated the ing social space alone, affecting both structural and dynamical communication. Hence, this data allows us to reconstruct the properties of social network. Our results offer new and signifi- daily trajectory of each mobile phone user for an extended pe- cant empirical evidence in connecting social networks and hu- riod of time. The dataset is massive and of excellent Longitu- man mobility. dinality: It covers activities of 10 million customers for more The remainder of this paper is organized as follows. than five years. In this study, we used data from two consecutive Section II provides detailed description of the mobile phone years. The database offers detailed empirical observation of hu- dataset. In Section III we explore the basic characteris- man mobility. For example, in Fig. 1 we show the three month tics of social networks and human mobility pattern, respec- trajectory of one users in our database, illustrating the nature tively. Section IV studies the co-locations between trajectories of the mobility patterns that can be extracted from the dataset. of pair of friends. Section V exams the proximity between hu- The different cell phone towers are denoted as grey dots, and man mobility and social networks in orders of increasing sophis- the Voronoi cell in grey marks the approximate reception area tication, from lower to higher order connections (a social tie to of each tower. Note that the trajectories are time resolved. That a local clustering). Sections VI investigates how human mobil- is, each time a user makes a call, the closest tower that routes ity impact underlying social network, for static and dynamical the call is recorded, serving as a proxy of the user’s appropriate properties respectively. Section VII concludes this paper. location at the moment. There are two caveats about the dataset that are particularly worth mentioning: (a) The user location is subject to the spa- II. DATA DESCRIPTION tial resolution of the mobile phone towers, which on average In the past few years, a collaboration with an European mo- results in uncertainties around 1-3 km. (b) User’s location is bile communication company grants us access to anonymized only recorded when the user uses the phone, hence we have no country wide mobile phone dataset, containing call records of knowledge of the user’s whereabouts between these active ses- their customers and collected for billing purposes. Being mobile sions. To cope with such caveats, improve temporal resolution communication records, the dataset naturally offers information of location sampling, and ensure our results are not affected by on wireless social communications through phone calls and text them, we constructed a second dataset that contains 50,000 in- messages between individuals. At the same time, whenever a dividuals, chosen from the 10 million mobile phone users based 102 JOURNAL OF COMMUNICATIONS AND NETWORKS, VOL. 17, NO. 2, APRIL 2015 100 100 10−1 10−1 10−2 10−3 10−2 10−4 P(k) P(d) 10−3 10−5 10−6 10−4 10−7 10−8 10−5 100 101 102 103 100 101 102 103 k d Fig. 2. Degree distribution of social network, where degree k measures the Fig.