A multi-layer model for diffusion of urgent information in mobile networks
Introduction
Today it is hard to imagine our life without cell phones. People make calls every day all around the world. Despite the fact that every person can call anyone, whose phone number he or she knows or has in the contact list, people often have a certain social circle − relatively small group of people, whom they call regularly [1]. Members of this group may differ by their roles and number from one person type to another: family members and some friends for the ordinary people; a lot of friends and colleagues for very communicative people; business partners and clients for the businessmen, etc. If we know social circle of a specific person, we might assume that we can quite accurately predict, whom this person will call in the particular situation. However, there are several factors that should be taken into consideration: cellular network and mobile device may have their own impact from the point of accessibility; social circle of the person is also dynamic, reflecting changes in the life state of the person (family, work, education, etc.). Taking all this together, we can say that the high-quality simulation of the calling activity is a very hard problem. But, using the call details records (CDR) provided by telecom companies, researchers are able to analyze the behavior of users in the mobile networks [2], [3] and provide viable information for building mobile networks models. Even more complicated and interesting problem is analyzing and simulation of the information diffusion in the cellular networks. This property cannot be directly obtained from the large mobile networks, because we cannot ask all participants whether or not did they say some information to the other user. That is why the only option for investigation of the information spreading on the large real networks is building the network of user-to-user interactions and usage of various spreading models on them [4], [5]. But when it comes to the analysis of the people behavior in terms of the calling activity in extreme situations, like flood, fire or terrorist attacks, we have even less options, because it is very hard to gather CDRs for the period of some extreme event [6]. All the above along with the need to accurately predict the consequences of the extreme events leads to the idea of the development of a flexible and precise model for the information diffusion simulation to analyze various urgent scenarios in the absence of the real data. In our previous work [7] we presented the agent-based multi-layer model for the information spreading in the different scenarios. The main contributions of this paper are: more thorough analysis of the available datasets for agents types identification; simplification of the model to eliminate numerous assumptions and arbitrary parameters; and comparison of the modelling results with the existing researches in the area of information spreading in the mobile networks.
Section snippets
Problem statement
Information diffusion in the mobile networks is a very complex process, which involves a great number of various aspects, from the hardware communications between mobile devices through the cell towers to interpersonal relations between people, who make calls and disseminate the information. All these aspects are connected one with another and create a sophisticated structure, which can be represented using layers (Fig. 1), where each layer stands for a specific aspect with its specific
Social networks analysis
Analysis of the information about the characteristics and evolution of social networks has got a lot of attention throughout years, as it can provide insights about the human behavioral patterns. The analysis of several datasets representing different types of the social networks, like the movie actors collaboration or citation network, was discussed in [3]. A network of phone calls made during August 10, 1998 was examined in this work. Dataset contains over 53 million nodes. Values of the
Model description
In our work for the processes of making calls and information transfer we used a multi-agent simulation due to its flexibility and the ability to accurately represent complex systems [26], [40], [41]. The core of this approach is a set of agent types, their characteristics and rules of agents’ interaction. Thus to create a proper multi-agent model, we need to describe agents’ parameters, identify several types of agents, define rules of their behavior and specify a number of global model
Urgent situation scenario
To investigate the information spreading process in the developed model for the urgent situations, like information about floods or terrorist attacks, we made significant changes into the calling process of the agents. On the step of selection the person from the contacts list to call, agents now select contacts by their call probability, which means that agents now call people, who are important for them, in the first place. Agents also remember the contacts they already called to and do not
Experiments
To compare the results produced by our model with the results obtained by other researchers in the area of information diffusion in the cellular networks, we conducted a set of experiments, where we varied the network size and measured the speed of the information spreading in the network. We refer to the paper by Karsai et al. [60], which is dedicated to the investigation of the SI spreading dynamics with simulations using the event sequences. The paper uses the same class of spreading models
Conclusion and future works
In this work we presented a multi-layer agent-based model for the information spreading in the mobile network. The main idea behind the concept of the model is to try to consider all levels involved into dissemination of the information − hardware levels (cell towers and mobile devices) and social interaction levels (contact network, calls network and information spreading model). In this work we took into account only social interactions levels, development and investigation of hardware levels
Acknowledgement
This paper is financially supported by The Russian Scientific Foundation, Agreement #14-11-00823 (15.07.2014)
Alexander Visheratin is a Ph.D. student at ITMO University (Saint Petersburg) since 2014, where he participates in the designing of workflow scheduling algorithms as well as in the development of big data processing systems. He got his specialist's Degree in informational systems in the ITMO University in 2014. Alexander's research interests are distributed computing, data mining and data analysis, and agent simulation of social processes.
References (61)
- et al.
Urgent information spreading multi-layer model for simulation in mobile networks
Procedia − Procedia Comput. Sci.
(2016) - et al.
The power of a good idea: quantitative modeling of the spread of ideas from epidemiological models
Phys. A Stat. Mech. Appl.
(2006) - et al.
Network robustness and irreversibility of information diffusion in Complex networks
J. Comput. Sci.
(2011) - et al.
A computational model for multi-agent E-commerce negotiations with adaptive negotiation behaviors
J. Comput. Sci.
(2013) Data clustering: 50 years beyond K-means
Pattern Recognit. Lett.
(2010)Silhouettes: a graphical aid to the interpretation and validation of cluster analysis
J. Comput. Appl. Math.
(1987)- et al.
Information spreading on dynamic social networks
Commun. Nonlinear Sci. Numer. Simul.
(2014) - et al.
Toolbox for visual explorative analysis of complex temporal multiscale contact networks dynamics in healthcare
Procedia Comput. Sci.
(2016) - et al.
Internet paradox: a social technology that reduces social involvement and psychological well-being?
Am. Psychol.
(1998) - et al.
The anatomy of urban social networks and its implications in the searchability problem
Sci. Rep.
(2015)
Statistical mechanics of complex networks
Rev. Mod. Phys.
Structure and tie strengths in mobile communication networks
Proc. Natl. Acad. Sci. U. S. A.
Multiscale analysis of spreading in a large communication network
J. Stat. Mech. Theory Exp.
Wired to connect: analyzing human communication and information sharing behavior during extreme events
Kdd-Lesi
A survey of results on mobile phone datasets analysis
EPJ Data Sci.
Analysis of a large-scale weighted network of one- to-one human communication
New J. Phys.
On the structural properties of massive telecom call graphs: findings and implications
Proc. 15th ACM Int. Conf. Inf. Knowl. Manag.
On random graphs
Publ. Math.
Exploring complex networks
Nature
Collective dynamics of ‘small-world’ networks
Nature
Emergence of scaling in random networks
Science (80-)
A random graph model for massive graphs
Proc. Thirty-second Annu. ACM Symp. Theory Comput. − STOC ’
An analysis of the optimum node density for ad hoc mobile networks
IEEE Int. Conf. Commun.
Vehicular ad hoc networks (VANETS): status, results, and challenges
Telecommun. Syst.
An epidemic model for information diffusion in MANETs
Proc. 5th ACM Int. Work. Model. Anal. Simul. Wirel. Mob. Syst. − MSWiM ’02
Information contagion: an empirical study of the spread of news on digg and twitter social networks
Proc. Fourth Int. AAAI Conf. Weblogs Soc. Media
Outtweeting the twitterers − predicting information cascades in microblogs
Proc. 3rd Conf. Online Soc. Networks
Complex-network modeling of a call network
IEEE Trans. Circuits Syst. I Regul. Pap.
The small world yields the most effective information spreading
New J. Phys.
Cited by (12)
An agent-based model to simulate inhabitants’ behavior during a flood event
2021, International Journal of Disaster Risk ReductionCitation Excerpt :The combination of flood and inhabitant behavior, is not the key element of these models. For instance Ref. [37], provided an ABM to simulate the information spreading in the mobile network. The model is not specifically dedicated to flood, but can be used facing this risk [20].
Social awareness of crisis events: A new perspective from social-physical network
2020, CitiesCitation Excerpt :In terms of social network topology, Huang, Shen, and Cheng (2016) proposed a super network model to describe the topology of public opinion network based on the theory of complex networks. Visheratin, Trofimenko, Mukhina, Nasonov, and Boukhanovsky (2017) proposed a time-delay evolution model in social network to analyze the interaction mechanism between two social events. As for community detection, which is used to identify meaningful subgroups within a complex network, also plays an important role in extracting the evolution information of crisis events from the social network.
A novel negative feedback information dissemination model based on online social network
2019, Physica A: Statistical Mechanics and its ApplicationsCitation Excerpt :The traditional network evolution analysis method has failed to meet the requirements. There is an urgent need for new models and analytics for large-scale social networks [11–15]. An important research direction of the social network evolution calculation method is to microscopically model the users of social networks.
Modeling information dissemination and evolution in time-varying online social network based on thermal diffusion motion
2018, Physica A: Statistical Mechanics and its ApplicationsData through the Computational Lens
2017, Journal of Computational ScienceCombining the Information Diffusion Model for the Propagation of Zhouyi Ideology in the New Era
2024, Applied Mathematics and Nonlinear Sciences
Alexander Visheratin is a Ph.D. student at ITMO University (Saint Petersburg) since 2014, where he participates in the designing of workflow scheduling algorithms as well as in the development of big data processing systems. He got his specialist's Degree in informational systems in the ITMO University in 2014. Alexander's research interests are distributed computing, data mining and data analysis, and agent simulation of social processes.
Tamara B. Trofimenko is a graduate of ITMO University. She got a double master’s degree at University of Amsterdam and ITMO University with a thesis dedicated to the development of agent-based information spreading model. Her major research interests focus on machine learning, data analysis and agent-based simulation.
Ksenia Mukhina is a Ph.D. student at ITMO University (Saint Petersburg). She received a M.S. in 2016 from ITMO University at Saint Petersburg with a thesis dedicated to visual analysis of dynamic processes in temporal networks. Her research interests are in the areas of complex network visualization, map visualization, and geographical information systems.
Denis Nasonov is a researcher at eScience Research Institute at ITMO University (Saint-Petersburg). He is leading infrastructure research group, which is focused on several fields, including distributed computation systems, cloud computing, optimization scheduling algorithms, data placement algorithms, meteorological simulation, complex network processes.
Alexander V. Boukhanovsky is the Chair of High Performance Computing (HPC) Department in ITMO University. In 2005, he defended his dissertation on Concurrent Software Statistical Measurements of Spatial-Temporal Fields. Since 2006 he has been working as a Professor of Information Systems and Head of the Parallel Software lab in ITMO University. In 2007 he created the eScience Research Institute. His research interests are high-performance computing, computer modelling of complex systems, intelligent computational technologies, distributed environments for multidisciplinary researches, statistical analysis and simulation in marine sciences.