Synthesizing Social Proximity Networks by Combining Subjective Surveys with Digital Traces Christopher Barrett*, Huadong Xia*, Jiangzhuo Chen*, Madhav.

Slides:



Advertisements
Similar presentations
Complex Networks for Representation and Characterization of Images For CS790g Project Bingdong Li 9/23/2009.
Advertisements

Mobile Communication Networks Vahid Mirjalili Department of Mechanical Engineering Department of Biochemistry & Molecular Biology.
Yang Yang, Miao Jin, Hongyi Wu Presenter: Buri Ban The Center for Advanced Computer Studies (CACS) University of Louisiana at Lafayette 3D Surface Localization.
Logistics Network Configuration
Power Laws By Cameron Megaw 3/11/2013. What is a Power Law?
Modularity and community structure in networks
Directional triadic closure and edge deletion mechanism induce asymmetry in directed edge properties.
Forwarding Redundancy in Opportunistic Mobile Networks: Investigation and Elimination Wei Gao 1, Qinghua Li 2 and Guohong Cao 3 1 The University of Tennessee,
University of Buffalo The State University of New York Spatiotemporal Data Mining on Networks Taehyong Kim Computer Science and Engineering State University.
Leting Wu Xiaowei Ying, Xintao Wu Dept. Software and Information Systems Univ. of N.C. – Charlotte Reconstruction from Randomized Graph via Low Rank Approximation.
SimDL: A Model Ontology Driven Digital Library for Simulation Systems Jonathan Leidig - Edward A. Fox Kevin Hall Madhav Marathe Henning Mortveit.
Ontology Classifications Acknowledgement Abstract Content from simulation systems is useful in defining domain ontologies. We describe a digital library.
Clustering short time series gene expression data Jason Ernst, Gerard J. Nau and Ziv Bar-Joseph BIOINFORMATICS, vol
© University of Minnesota Data Mining for the Discovery of Ocean Climate Indices 1 CSci 8980: Data Mining (Fall 2002) Vipin Kumar Army High Performance.
Sampling from Large Graphs. Motivation Our purpose is to analyze and model social networks –An online social network graph is composed of millions of.
Community Detection in a Large Real-World Social Network Karsten Steinhaeuser Nitesh V. Chawla DIAL Research Group University of Notre.
Clustering Unsupervised learning Generating “classes”
Chapter 12: Analysis of Variance
The Erdös-Rényi models
Information Networks Power Laws and Network Models Lecture 3.
Social Network Analysis via Factor Graph Model
Computational Methods for Testing Adequacy and Quality of Massive Synthetic Proximity Social Networks Huadong Xia, Christopher Barrett, Jiangzhuo Chen,
Chapter 11: Artificial Intelligence
Spectral coordinate of node u is its location in the k -dimensional spectral space: Spectral coordinates: The i ’th component of the spectral coordinate.
FAEIS Project User Opinion Survey 2005 Thursday, June 23, 2005 Washington, D.C. H. Dean SutphinYasamin Miller ProfessorDirector, SRI Agriculture & Extension.
UNIVERSITY of NOTRE DAME COLLEGE of ENGINEERING Preserving Location Privacy on the Release of Large-scale Mobility Data Xueheng Hu, Aaron D. Striegel Department.
1 / 27 California Educational Research Association 88 th Annual Conference Formative Assessment: Implications for Student Learning San Francisco, CA November.
Dynamic Clustering for Acoustic Target Tracking in Wireless Sensor Network Wei-Peng Chen, Jennifer C. Hou, Lui Sha.
Wei Gao1 and Qinghua Li2 1The University of Tennessee, Knoxville
Information Flow using Edge Stress Factor Communities Extraction from Graphs Implied by an Instant Messages Corpus Franco Salvetti University of Colorado.
Comparing Effectiveness of Top- Down and Bottom-Up Strategies in Containing Influenza Achla Marathe, Bryan Lewis, Christopher Barrett, Jiangzhuo Chen,
A Graph-based Friend Recommendation System Using Genetic Algorithm
EpiFast: A Fast Algorithm for Large Scale Realistic Epidemic Simulations on Distributed Memory Systems Keith R. Bisset, Jiangzhuo Chen, Xizhou Feng, V.S.
Testing Models on Simulated Data Presented at the Casualty Loss Reserve Seminar September 19, 2008 Glenn Meyers, FCAS, PhD ISO Innovative Analytics.
Aemen Lodhi (Georgia Tech) Amogh Dhamdhere (CAIDA)
Emergence of Scaling and Assortative Mixing by Altruism Li Ping The Hong Kong PolyU
Workshop on Applications of Wireless Communications (WAWC 2008) 21 August 2008, Lappeenranta - Finland CONTROLLING EPIDEMICS IN WIRELESS NETWORKS Ranjan.
Showcase /06/2005 Towards Computational Epidemiology Using Stochastic Cellular Automata in Modeling Spread of Diseases Sangeeta Venkatachalam, Armin.
Network Dynamics and Simulation Science Laboratory Cyberinfrastructure for Social Networks and Network Analysis Network Dynamics and Simulation Science.
Xiaowei Ying, Xintao Wu Dept. Software and Information Systems Univ. of N.C. – Charlotte 2008 SIAM Conference on Data Mining, April 25 th Atlanta, Georgia.
Scaling Agent-based Simulation of Contagion Diffusion over Dynamic Networks on Petascale Machines Keith Bisset Jae-Seung Yeom, Ashwin Aji
Simulating Diffusion Processes on Very Large Complex networks Joint work with Keith Bisset, Xizhou Feng, Madhav Marathe, and Anil Vullikanti Jiangzhuo.
Coevolution of Epidemics, Social Networks, and Individual Behavior: A Case Study Joint work with Achla Marathe, and Madhav Marathe Jiangzhuo Chen Network.
Network Community Behavior to Infer Human Activities.
Performance of Adaptive Beam Nulling in Multihop Ad Hoc Networks Under Jamming Suman Bhunia, Vahid Behzadan, Paulo Alexandre Regis, Shamik Sengupta.
1. Background A generic ‘chirp’ can be closely approximated by a connected set of multiscale chirplets with quadratically-evolving phase. The problem of.
PCB 3043L - General Ecology Data Analysis.
Comparison of Tarry’s Algorithm and Awerbuch’s Algorithm CS 6/73201 Advanced Operating System Presentation by: Sanjitkumar Patel.
Brief Announcement : Measuring Robustness of Superpeer Topologies Niloy Ganguly Department of Computer Science & Engineering Indian Institute of Technology,
Optimal Interventions in Infectious Disease Epidemics: A Simulation Methodology Jiangzhuo Chen Network Dynamics & Simulation Science Laboratory INFORMS.
1 Finding Spread Blockers in Dynamic Networks (SNAKDD08)Habiba, Yintao Yu, Tanya Y., Berger-Wolf, Jared Saia Speaker: Hsu, Yu-wen Advisor: Dr. Koh, Jia-Ling.
Global Clock Synchronization in Sensor Networks Qun Li, Member, IEEE, and Daniela Rus, Member, IEEE IEEE Transactions on Computers 2006 Chien-Ku Lai.
Comparison of Individual Behavioral Interventions and Public Mitigation Strategies for Containing Influenza Epidemic Joint work with Chris Barrett, Stephen.
Community structure in graphs Santo Fortunato. More links “inside” than “outside” Graphs are “sparse” “Communities”
Fast Parallel Algorithms for Edge-Switching to Achieve a Target Visit Rate in Heterogeneous Graphs Maleq Khan September 9, 2014 Joint work with: Hasanuzzaman.
Network Partition –Finding modules of the network. Graph Clustering –Partition graphs according to the connectivity. –Nodes within a cluster is highly.
James Hipp Senior, Clemson University.  Graph Representation G = (V, E) V = Set of Vertices E = Set of Edges  Adjacency Matrix  No Self-Inclusion (i.
Efficient Implementation of Complex Interventions in Large Scale Epidemic Simulations Network Dynamics & Simulation Science Laboratory Jiangzhuo Chen Joint.
Network Dynamics and Simulation Science Laboratory Structural Analysis of Electrical Networks Jiangzhuo Chen Joint work with Karla Atkins, V. S. Anil Kumar,
A Place-based Model for the Internet Topology Xiaotao Cai Victor T.-S. Shi William Perrizo NDSU {Xiaotao.cai, Victor.shi,
Epidemic spreading on preferred degree adaptive networks Shivakumar Jolad, Wenjia Liu, R. K. P. Zia and Beate Schmittmann Department of Physics, Virginia.
Network Science in NDSSL at Virginia Tech
PCB 3043L - General Ecology Data Analysis.
Network Science: A Short Introduction i3 Workshop
Wireless Epidemic The wireless epidemic (Nature 449, ; 2007) by Jon Kleinberg ‘Digital traffic flows not only over the wired backbone of the Internet,
Department of Computer Science University of York
复杂网络可控性 研究进展 汪秉宏 2014 北京 网络科学论坛.
The Impact of Changes in Network Structure on Diffusion of Warnings
Presentation transcript:

Synthesizing Social Proximity Networks by Combining Subjective Surveys with Digital Traces Christopher Barrett*, Huadong Xia*, Jiangzhuo Chen*, Madhav V. Marathe*, Henning S. Mortveit* and Marcel Salathe # * The Network Dynamics and Simulation Science Laboratory, Virginia Bioinformatics Institute at Virginia Tech, USA # Center for Infectious Disease Dynamics, Penn State University, USA IEEE CSE2013

We thank our external collaborators and members of the Network Dynamics and Simulation Science Laboratory (NDSSL) for their suggestions and comments. This work has been partially supported by DTRA Grant HDTRA , DTRA CNIMS Contract HDTRA1-11-D , NIH MIDAS Grant 2U01GM , NSF PetaApps Grant OCI , NSF NetSE Grant CNS Acknowledgement

Close proximity relations between people are critical in understanding the diffusion of influenza-like epidemics. Those close proximity relations are modeled collectively as a social contact network. Existing solutions in constructing social contact networks: – Digital devices to detect proximity between people: RFID tags, cell phones, motes, etc. – Subjective assessment and survey information Background: Model Close Proximity Relations Between People Modeling Social contact network Social contact network

Solution 1: Digital Devices to Detect Proximity Between People Free of human error High quality Expensive Nontrivial to generalize 700-student contact Network => 1000-student contact Network?

Solution 2: Subjective Assessment and Survey Information Complete Graph G(n,p) Geometry Random Graph Subjective Assessment … Inexpensive Easy to generalize Sublocation interactions remains a black box

A hybrid methodology that combines both subjective surveys and digital traces: – Generic pattern exists in a very small location: conference room, class room, restaurant at different hours. As a Showcase: School networks New Solution: A Hybrid Methodology

Data sets Objective 1: understand In-class contact networks – Identifying class intervals – Extracting class networks Objective 2: generative network model that model the digital trace network Objective 3: comparison study on the impact of detailed sublocation structure Outline

Populations: – NRV population: 150K – High school population: 2.5K We collected class schedules for 3 schools in New River Valley Region Data Sets: Surveys

Digital trace data – Collected from an American high school – 788 participants, including 655 students, 73 teachers and 55 staff members, and 5 other people (94% of the school population) – Each participant carry a mote for an entire typical school day. – Their motes detect other motes within 3 meters for every 20 seconds, stored as CPRs in the data set CPR: close proximity records CPI: close proximity interaction, a continuous sequence of CPRs Contacts: a contact is the sum of all CPIs between two motes. – 2,148,991 CPRs, 762,868 CPIs and 118,291 contacts Data Sets: Digital Trace Data

Data sets Objective 1: understand In-class contact networks – Identifying class intervals – Extracting class networks Objective 2: generative network model that model the digital trace network Objective 3: comparison study on the impact of detailed sublocation structure Outline

Formation of school networks: Step to identify class networks: – Identify class periods – For each identified class period, identify within-class contact networks Structure of School Networks

Motes are anonymized and the class schedules are unknown. Mote Signals are highly volatile – Directional – Unstable device Challenges (1)

Classes and Breaks Reveal Quite Different Patterns

Use the Algorithm to Plot Time Zone for Class Breaks

Challenges (2): Isolate In-Class Contact Networks Interference exists for sensor Signals! – A very large Connected Component for any snapshot contact networks – Sensor Signals can traverse the wall (via windows/doors)?

Isolate In-Class Contact Networks CPIs within the same class interval comprise a relative stable contact network, even if CPIs are volatile --- foundation for us to analyze CPIs traverse across classrooms but we hypothesize: – CPIs between classrooms are short and unstable  An “test and try” algorithm to remove noises – CPIs between classrooms are sparser than within  Modularity based Community Detection Algorithm

Detect School Communities: Modularity Based Algorithm

 Alternative Slide

Students in the class typically form into one or multiple groups; students of the same group are highly connected. Duration of CPIs follow a power law like distribution Analyze In-class Contact Network 47 nodes 21 nodes 32 nodes

Data sets Objective 1: understand In-class contact networks – Identifying class intervals – Extracting class networks Objective 2: generative network model that model the digital trace network Objective 3: comparison study on the impact of detailed sublocation structure Outline

G(n,p) model is not appropriate: – Cannot: match degree, match clustering coefficients – Can: match n; match the sum of edge weights by adjusting p Chung-Lu model: match both degrees and edge weights – List of degree k v of each node v from a digital trace template – Chung-Lu model connect each node pair (v, u) with probability where m is the total edge number – We adjust the edge weight for each generated edge, so that the edge weight follow a power law distribution. ERGM model: – more powerful candidate – complex compared to Chung-Lu model Use Theoretic Graph Models to Fit Digital Trace Templates

Spectral Gap of a network: the difference between the largest two eigenvalues of the network adjacency matrix A larger spectral gap means the disease is easier to spread on the network. Compare Spectral Gaps between Digital Trace Templates and Graph Models

Data sets Objective 1: understand In-class contact networks – Identifying class intervals – Extracting class networks Objective 2: generative network model that model the digital trace network Objective 3: comparison study on the impact of detailed sublocation structures Outline

Aim: To compare three in-class models within a realistic context, we use the three models to construct three types of high school networks, and further embed school networks within the larger regional network Input: – High school populations in NRV region – The NRV population in NRV regions Output: – Three types school networks based on three in-class models respectively – Three types of NRV Network based on three in-class models respectively School Networks and the Region Network

The school network based on calibrated ChungLu model seems a good match to that based on digital trace templates, structurally. Structural Properties of School Networks Embedded with Different In-class Models

Disease Spread in a Social Network Within-host disease model: SEIR Between-host disease model: – probabilistic transmissions along edges of social contact network – from infectious people to susceptible people

Simulation to ILI without Intervention Vaccine high degree nodes Vaccine high degree nodes +social distance Epidemic Dynamics of School Networks Embedded with Different In-class Models

ANOVA peakday Sum of SquaresdfMean SquareFSignificance Between Groups * Within Groups Total Epicurve Difference with Different In-class Models Multiple Comparisons Dependent Variable: peakday Tukey HSD (I) groups(J) groupsMean Difference (I-J)Significance G(n,p) Digital trace *.022* ChungLu Digital trace G(n,p) *.022* ChungLu ChungLu G(n,p) Digital trace *. The mean difference is significant at the 0.05 level.

The digital trace based templates capture network structures that are critical in understanding the role of interventions, and not available in previous theoretic sublocation models such as G(n,p) It is possible to capture a faithful structural features or dynamics by tuning appropriate theoretic graph models like Chung-Lu to the real digital trace templates, at least under some limited scenarios. ERGM could possible serve as a good model, but Chung-Lu model seems like a reasonable fit for now. Summary of the Comparison Study

We show a hybrid methodology that combines subjective survey with digital trace data. In-class contact structure is important in understanding epidemics and intervention strategies. Our methodology is generic, applicable to other template networks – Office building – Military bases – Hospital rooms – … … Conclusions

Questions?

Extra slides

Similarity between Community Division

Types of classroom organization: teacher-centered or peer-based (internet source: Research Unit for Multilingualism and Cross-Cultural Communication)Research Unit for Multilingualism and Cross-Cultural Communication Illustration to Class Network Topology Structure

Construction of a High School Network

Embed School Networks Within a Larger Regional Network