Elsevier

Neurocomputing

Volume 174, Part B, 22 January 2016, Pages 605-616
Neurocomputing

Mining community and inferring friendship in mobile social networks

https://doi.org/10.1016/j.neucom.2015.09.070Get rights and content

Abstract

Along with the rapidly growth of mobile terminals and wireless technologies, mobile social networking services are very popular with peoples. Recently many mobile social platforms based on location-based service are developed to allow users to share their check-ins and events with friends. Check-ins data in location-based mobile social networks as well as call detail records (CDR) in mobile communication network may provide insight into community structure, relationships and members in the network. In this paper, we study the problem of community detection and friendship prediction in mobile social networks. We have presented a method to find community structure built on combination entropy, and evaluate modularity of a virtual campus mobile network (V-Net). The outcomes demonstrate that the proposed algorithm mine meaningful communities according to users׳ registration. We investigate the potential friendship among users by taking into account both users׳ links with friends and their check-ins at various positions in Gowalla. This work describes the probability distributions of friendships per number of friends, number of check-ins and number of visited places. The findings confirm that our approaches achieve well performance with aggregated features of user similarity and place entropy than other methods. Moreover, members reveal different social properties in the two networks, in the V-Net influence users tend to hold community together, while in Gowalla community members are likely to visit the common positions.

Introduction

The growing number of rich sensors and communication enabled devices brings opportunities for perceiving the environment in innovative ways. Nowadays smartphone not only serves as the key communication and computing mobile terminate of choice, but it also equips with an abundant set of embedded sensors including a camera, microphone, accelerometer, gyroscope, digital compass, and GPS. Generally, these sensors promote new applications across a wide variety of fields, such as social networking services [1], [2], transportation [3], [4], user mobility [5], [6], touring routes [7], [8], and business sites selection [9], [10]. The most popular of those applications is location-based services platform in the mobile social networks.

Location-based social networks (LBSNs) become more popular and excite great interest among people, who are addressing daily opinions on each others web spaces, having a chat, running a blog, writing comments, posting and noticing pictures and videos, preparing activities and last but not least, updating their current status. With ubiquitous mobile devices including smartphones and GPS devices, social networking platforms built on location-based services, such as Facebook Places, Foursquare, Gowalla, and Brightkite, become more and more popular among teenagers. For an instance, as of January, 2014, Foursquare claims to have over 45 million people worldwide and over 5 billions of check-ins with millions more every day. Users of a location-based social networking platform can pick their friends list and obtain listed as friends to other users as well as traditional social network [11], [12], [13]. They can check in and share their positions and events to their friends via the social network. Furthermore, users may alert friends to their check-ins when visiting a venue (e.g., museum, hostel, and shopping mall) via their mobile phones [14], [15], [16]. Users are encouraged to check in at position to earn badges, venue mayorships and receive special offers.

Community detection and friendship prediction is a key topic in social networking field that provides special approach whereby the study can benefit scientists as well as businesses in a variety of domains. The community structure is an important property which can disclose several unknown characteristics for a given network. Members belong to the same group or community are possible to have common interests or similar properties. Social networks can represent a variety of relationships including friendship, kinship and ties among the participants. Zheleva investigated the predictive power of overlaying friendships and family ties on three interesting social networks in the real world [17]. Users can push and show their outdoor activities including touring and cheering with other users in LBSNs, and they can conveniently manage their footprints and also share them with friends. Ref. [18] studied the correlation between friendship and users׳ attributes, such as their mobility features, social graph properties, and users׳ profiles in a commercial LBSNs. A potential friendship links among individuals can be predicted via their common friends, interests and visited places in mobile social network. Social structures tend to be highly dynamic objects because they develop and vary rapidly over time via adding new nodes or edges, so the friendship link prediction problem become more challenging.

In this paper, we have presented a measurement study of mining community detection and inferring friendship on two different datasets generated by users׳ cellphone in mobile social network. The cellphone communication dataset is a virtual campus mobile network (V-Net) provided by a major mobile communication service corporation in China. Users in the V-Net are students or faculties who are from a college or university, and they could dial with each other more conveniently and cheaply via virtual short phone number. We extract individuals׳ call detail records (CDR) during 3 months along with their registration information. The second dataset is an online location-based social network, namely Gowalla, which is a location-based service platform launched in 2007 and shut down in 2012. Users are able to check in at venues in their local proximity via a special application on mobile devices.

Our main contributions can be summarized as follows:

(1) We propose an algorithm for finding community and evaluating modularity of the virtual campus mobile network (V-Net). We find that community structure is influenced by important nodes, edge density and internal degree (see Section 3). The number of communities in the network vary the combination entropy threshold (see Section 5).

(2) We investigate potential friend links by analyzing not only users׳ friends of friends, but also places visited by users׳ friends in Gowalla (see Section 4). We describe the probability distributions of friendships per number of friends, number of check-ins and number of places of users respectively, those follow log-normal distribution (see Section 5).

(3) We study the friendship prediction by taking into account place properties and user similarity (see Section 4). The average opportunity for a couple of individuals who have visited at the common positions may be friends drops while the entropy of the position increases. There exits strong correlative between friendship probability and user similarity according to the growing trend of the curvy line (see Section 5).

Our findings uncover that users reveal different social activity in the two mobile social network, e.g., in the V-Net some key users tend to promote friendship and arise community, however, in Gowalla community members are likely to visit the common positions and form friendship around the places.

Section snippets

Related worked

Social networks such as Facebook, Foursquare, Weibo, and Momo have aroused interest of millions of users, some of these social platforms have grew rapidly more than three quarters in the past year. Recent statistics reports present that social networking services have overcome search engines in terms of application. eMarketer [19] addresses that it expects 4.55 billion people worldwide to use a mobile phone in 2014, and mobile phone penetration will rise from 61.1% to 69.4% of the global

Detecting community in mobile social network

Typically the investigation of community structure in networks is actually a really important issue in many fields as well as disciplines. This main problem is associated with social network analyzing tasks, biological inquiries and technological issues, successively for example, objective analysis of relationships on the Internet, functional studies in metabolic and protein networks, or optimization of large infrastructures. Several social networks have a basic property of including community

Inferring friendship in location-based mobile social network

In social networks, individuals are linked with relationships and these connections form the basic structure of their community of the network. A community which referred to a module or a cluster is generally considered as a collection of nodes with many more ties amongst its members than the remainder individuals of the network. Many real world problems can be effectively modeled as complex relationship networks where nodes represent entities of interest and edges mimic the interactions or

Experimental

We currently present the experimental outcome and analysis of our methods proposed in this paper. This section conducts an investigation of detecting community from call detailed records (CDR) log and inferring friendship from check-in records with location-based service in mobile social network. The experimental results disclose that the intercall durations follow a power-law distribution with an exponential and common interests or interactions between members form communities, specially, some

Conclusion

In this paper, we investigate community detection and friendship prediction on two different datasets from mobile social network. We have presented a measurement to mine community structure and to evaluate modularity of the V-Net, of which users register their social roles in university. Our outcomes confirm that the proposed algorithm discovers meaningful communities compare to student groups in real society. The other dataset is an online location-based social network, namely Gowalla, and we

Acknowledgments

This research is supported by the Special Fund for Earthquake Research in the Public Interest no. 201508025, and the National Science and Technology Project of the Ministry of Education of China (No. MCM20121061, MCM20121041).

Ke Xu received the M.S. degrees from Huazhong University of Science and Technology in 2007. During 2007–2010, he stayed in South-Central University for Nationalities. He is now a Ph.D Candidate in the School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan, China. His research interests include data mining and social network analyze.

References (33)

  • Y. Zhang et al.

    Inertial sensors supported visual descriptors encoding and geometric verification for mobile visual location recognition applications

    Signal Processing

    (2015)
  • Y. Luo et al.

    Fast terrain mapping from low altitude digital imagery

    Neurocomputing

    (2015)
  • S. Scellato, A. Noulas, C. Mascolo, Exploiting place features in link prediction on location-based social networks, in:...
  • T. Guan et al.

    On-device mobile visual location recognition by integrating vision and inertial sensors

    IEEE Trans. Multimed.

    (2013)
  • S. Scellato, A. Noulas, R. Lambiotte, C. Mascolo, Socio-spatial properties of online location-based social networks,...
  • Y. Gao et al.

    3D Object retrieval and recognition with hypergraph analysis

    IEEE Trans. Image Processing

    (2012)
  • A. Noulas, S. Scellato, N. Lathia, C. Mascolo, Mining user mobility features for next place prediction in...
  • T. Guan et al.

    Efficient bof generation and compression for on-device mobile visual location recognition

    IEEE Multimed.

    (2014)
  • J. Cranshaw, E. Toch, J. Hong, A. Kittur, N. Sadeh, Bridging the gap between physical location and online social...
  • B. Wei et al.

    Projected residual vector quantization for ann search

    IEEE Multimed.

    (2014)
  • Y. Gao et al.

    Visual–textual joint relevance learning for tag-based social image search

    IEEE Trans. Image Process.

    (2013)
  • Ryan N. Lichtenwalter, Jake T. Lussier, Nitesh V. Chawla, New perspectives and methods in link prediction, in:...
  • T. Guan et al.

    On-device mobile visual location recognition by using panoramic images and compressed sensing based visual descriptors

    PLoS One

    (2014)
  • R. Ji et al.

    Location discriminative vocabulary coding for mobile landmark search

    Int. J. Comput. Vis.

    (2012)
  • Z. Cheng, J. Caverlee, K. Lee, D. Sui, Exploring millions of footprints in location sharing services, in: Proceedings...
  • R. Ji et al.

    Task-dependent visual-codebook compression image processing

    IEEE Trans. Image Processing

    (2012)
  • Cited by (43)

    • Social media data analytics using feature engineering

      2020, Systems Simulation and Modeling for Cloud Computing and Big Data Applications
    • Recommending similar users using moving patterns in mobile social networks

      2019, Computers and Electrical Engineering
      Citation Excerpt :

      Arb et al. [24] distinguishes users through phone numbers and discriminates a friend by comparing the near users’ phone numbers through a wireless technology over short distances [25] discriminates the similar users by considering the number of meetings and time in the same location based on mobile sensor data. Xu [18] takes into account friend types and check-in records to inferring the friendship between users in mobile social networks. [22] analyzes the user's behavior log such as time, space, and interest.

    View all citing articles on Scopus

    Ke Xu received the M.S. degrees from Huazhong University of Science and Technology in 2007. During 2007–2010, he stayed in South-Central University for Nationalities. He is now a Ph.D Candidate in the School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan, China. His research interests include data mining and social network analyze.

    Keju Zou received his BS degree from Sichuan University in 2008. He is currently a Ph.D. candidate at School of Engineering of Sun Yat-Sen University, Guangzhou, China. His research interests include sleep medicine and biomedical signal processing.

    Yan Huang received his BS degree from South-Central Univercity for Nationalities in 2012. He is currently a Ph.D Candidate in School of Computer Science and Technology, Huazhong University of Science and Technology. His research interests are machine learning, pattern recognition and knowledge base.

    Xiaoyang Yu received his Master degree from The University of Edinburgh for Artificial Intelligence in 2008. He was admitted to Huazhong University of Science and Technology in 2011 and is currently a Ph.D Candidate. His research interests are distributed computing and storage system.

    Xinfang Zhang received the Ph.D. degree from Huazhong University of Science and Technology in 1993. Now he is a Professor in the School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan, China. His research interests include data mining and cloud computing.

    View full text