Exploring User Social Behavior in Mobile Social Applications Konglin Zhu *, Pan Hui $, Yang Chen *, Xiaoming Fu *, Wenzhong Li + * University of Goettingen, $ Deutsche Telekom, + Nanjing University
Outline Background Motivation Mobile social application (MS app): Goose Experiment methodology User behavior analysis Information propagation in MS apps Conclusions 2
Background Mobile devices are increasing –1.2 billion mobile phones are sold in 2009 –There are around 5 billion mobile phone subscriptions worldwide Mobile social applications are popular –Mobile version of Facebook, Twitter April 2010: 62% mobile users in Twitter Facebook has 250 million mobile users –Location based mobile social applications Foursquare –Non-Internet social applications PeopleNet [1], Prism [2], Goose [3], … 3
Motivations Lack of knowledge on mobile social user behavior –Previous mobile devices deployment only investigates the user encounters, but no interactions Information propagation in mobile social networks –Most previous information propagation models are simulation based, no real deployment Our objectives: –Understand user behavior in mobile social applications User overall behavior User social behavior –Investigate information propagation in mobile social networks DTN routing efficiency Information epidemics 4
Mobile social application: Goose Goose –A mobile social application implemented on Nokia Symbian system Function of Goose –Exchange contact Exchange and update user profiles –Update status Post new status on the Goose wall –Message Unicast message via Bluetooth or SMS Broadcast message via Bluetooth –Search friends Search a specific friend from other friends contact lists 5
Experiment methodology Deployment –We deploy our software in two campuses 12 volunteers in University of Goettingen 15 volunteers in Nanjing University –The experiments last 15 days in each campus Data collection –Bluetooth MAC address –The time duration users run Goose –Cellular ID, nearby devices (every 2 minutes) –Incoming and outgoing events Message ID, message type, time received, sender, previous relays, message size 6
User behavior analysis User overall behavior –User activity –User sessions –User mobility –Message statistics User social behavior –User encounters –User interactions 7
User overall behavior (1) User activity –Active user is the user active at a certain time –It shows the periodicity bursts of active users (a) User activity in NJU (b) User activity in UGoe 8
User overall behavior (2) User sessions –A session is the time difference between switching on and switching off Goose –It reflects the frequency of using Goose 9 (a) User sessions in NJU (a) User sessions in UGoe
User overall behavior (3) User mobility on campus –Trace a users mobility by recording cellular ID –A typical users time duration on each cellular User time duration in each cellular 10
User overall behavior (4) Message statistics –Communication messages are more than other messages –UGoe has more event types than NJU Message statistics 11
User social behavior (1) Heavy tail of User encounters –Heavy tail distribution [4] It is known as scale-free network, it has been observed in many complex networks, such as Internet, WWW, sending –The number of encounters in a day by each user Encounters distribution in UGoe Encounters distribution in NJU 12
User social behavior (2) Pareto principles of user interaction –Pareto principle Known as rule: 80% of the effects comes from 20% of causes –Both encounters and interactions show Pareto principle –More encounters suggest more interactions between users User interactions vs. user encounters 13
User social behavior (2) Pareto principles of user interaction –Pareto principle Known as rule: 80% of the effects comes from 20% of causes –Both encounters and interactions shows Pareto principle –More encounters suggests more interactions between users User interactions vs. user encounters 14 Pareto principle of user interactions
Information propagation in MS apps (1) Small world phenomenon [5] –The distance between two people is within 6 hops –Most of messages are sent to destination within 6 hops 15 Relays of messages
Information propagation in MS apps (2) DTN routing efficiency –Goose uses Bubble Rap [6] as the routing strategy for message forwarding Forward the message based on the popularity of nodes –It shows the number of messages sent and received Delays of messages 16 StatusUnicastBroadcast Messages sent Total message received Unique messages received Messages sent vs. messages received
Information propagation in MS apps (3) 17 Message delays –Varies from 0 minutes to 10,000 minutes –Unicast messages have shorter delay than broadcast and status updates Delays of messages
Information epidemics [7] –Susceptible-Infectious- Susceptible Each node can be: –Susceptible –Infectious An infectious node can infect others with λ –We initialized an epidemic message in one device in UGoe –The infectious scale reach 50% in a short term, and 80% in the long run Information epidemics 18 Information propagation in MS apps (4)
Conclusions We study the user overall behavior and find that the user activity is similar as human work pattern We explore the user social behavior in which the user encounters follows a heavy tail distribution and user interactions follows Pareto principle We demonstrate the information propagation efficiency by DTN routing and information epidemics model in mobile social networks We expect to extend the function of Goose and have a larger size of deployment 19
References [1] M. Motani, V. Srinivasan, and P. Nuggehalli, PeopleNet: Engineering a Wireless Virtual Social Network. In Proc. of MobiCom, 2005 [2] T. Das, P. Mohan, V. N. Padmanabhan, R. Ramjee and A. Sharma, PRISM: Platform for RemoteSensing using Mobile Smartphones. MobiSys [3] N. V. Rodriguez, P. Hui and J. Crowcroft, Has Anyone Seen My Goose? Social Network Services in Developing Regions. CSE 2009 [4] A. L. Barabasi, The Origin of Bursts and Heavy Tails in Human Dynamics. Nature [5] D. J. Watts and S. H. Strogatz, Collective Dynamics of Small-world Networks, Nature [6] P. Hui, J. Crowcroft, and E. Yoneki. Bubble rap: Social-based forwardingin delay tolerant networks. MobiHoc [7] A. Chaintreau, P. Hui, J. Crowcroft, C. Diot, R. Gass and J. Scott, Impact of human mobility on opportunistic forwarding algorithms. IEEE Trans. Mob. Comp,
Questions? 21