| NSF Org: |
DMS Division Of Mathematical Sciences |
| Recipient: |
|
| Initial Amendment Date: | August 6, 2018 |
| Latest Amendment Date: | July 31, 2019 |
| Award Number: | 1830547 |
| Award Instrument: | Continuing Grant |
| Program Manager: |
Leland Jameson
DMS Division Of Mathematical Sciences MPS Direct For Mathematical & Physical Scien |
| Start Date: | August 1, 2018 |
| End Date: | December 31, 2021 (Estimated) |
| Total Intended Award Amount: | $141,685.00 |
| Total Awarded Amount to Date: | $141,685.00 |
| Funds Obligated to Date: |
FY 2019 = $105,731.00 |
| History of Investigator: |
|
| Recipient Sponsored Research Office: |
1960 KENNY RD COLUMBUS OH US 43210-1016 (614)688-8735 |
| Sponsor Congressional District: |
|
| Primary Place of Performance: |
1958 Neil Ave Columbus OH US 43210-1016 |
| Primary Place of Performance Congressional District: |
|
| Unique Entity Identifier (UEI): |
|
| Parent UEI: |
|
| NSF Program(s): | ATD-Algorithms for Threat Dete |
| Primary Program Source: |
01001920DB NSF RESEARCH & RELATED ACTIVIT |
| Program Reference Code(s): |
|
| Program Element Code(s): |
|
| Award Agency Code: | 4900 |
| Fund Agency Code: | 4900 |
| Assistance Listing Number(s): | 47.049 |
ABSTRACT
Modeling and analyzing spatially-determined and time-varying (spatiotemporal) interactions is at the forefront of research in many scientific and engineering disciplines, including the social and behavioral sciences, transportation, healthcare, economics, and epidemiology. This project represents spatiotemporal interactions of entities as a dynamic complex network and aims to develop statistically-principled methods for modeling, analyzing, and monitoring the dynamic interactions. The methods developed in this work will provide scalable solutions for problems relevant to threat detection, including understanding spreading of diseases and viruses through human proximity networks, understanding human migration patterns through geo-tagged social media data, and monitoring multi-modal urban mobility networks through video footage and sensor logs in a smart city. Graduate and undergraduate students will be trained in interdisciplinary data science through involvement in the research. New data structures, models, and algorithms for manipulating and analyzing spatiotemporal networks will be implemented in the widely-used NetworkX Python package.
The project aims to advance the field of spatiotemporal network analysis by developing new models and methods for representing, monitoring, and predicting spatiotemporal interactions. The research introduces new problem formulations, new analytical methods, and new algorithmic techniques for implementation. This project has three primary aims. First, the project will develop a dynamic embedding model in a latent hyperbolic space to represent spatiotemporal networks. This model enables tracking topological changes both at the network level and at the level of pairs of entities over time. Next, the project will investigate a network surveillance framework based on a multi-resolution exponential random graph model to monitor complex spatiotemporal systems for real-time anomalies and threats. Third, the project will develop a multivariate point process on collections of actors in a spatiotemporal network to model timestamped directed events across different regions in space. This project seeks to create an integrated framework for simultaneously monitoring systematic risk and detecting imminent threat to a system using multi-modal network monitoring techniques. The techniques under development will be utilized to monitor complex systems arising from massive spatiotemporal data accumulation, including data on human contacts through physical proximity, social media data, and event data such as homicides in city neighborhoods and conflicts between countries. The fundamental results derived in this work will guide research in modeling and inference on dynamic networks and will serve as a benchmark for future work.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
PUBLICATIONS PRODUCED AS A RESULT OF THIS RESEARCH
Note:
When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
Some links on this page may take you to non-federal websites. Their policies may differ from this site.
PROJECT OUTCOMES REPORT
Disclaimer
This Project Outcomes Report for the General Public is displayed verbatim as submitted by the Principal Investigator (PI) for this award. Any opinions, findings, and conclusions or recommendations expressed in this Report are those of the PI and do not necessarily reflect the views of the National Science Foundation; NSF has not approved or endorsed its content.
Supported by this grant, we developed statistical models, methods and theories for analysis of network connected spatiotemporal data. Our developments include:
(1) The Community Hawkes Independent Pairs (CHIP) model for analyzing relational events data. We developed a highly scalable estimation approach that can easily fit networks with hundreds of thousands of nodes and millions of events, and yet has rigorous statistical performance guarantees.
(2) The multivariate community Hawkes model improves the CHIP model and is capable of modeling realistic dependence patterns that are commonly observed in relational events data. However, fitting this model is not as computationally efficient as CHIP.
(3) The latent space Hawkes model which produces interpretable results and fit many salient features of several real data sets well.
(4) A joint latent space model for a social network and multivariate attributes that allows researchers to study student’s responses collected from a survey and their interactions with their peers simultaneously.
(5) A method for estimating causal network influence from observational data adjusting for latent homophily. Identifying the influence network-connected neighbors exert on individuals' outcomes is a fundamental scientific problem associated with network-linked data and our methods represent a significant advancement in this literature.
(6) Models and a statistical hypothesis testing framework for detecting and evaluating the small-world property that is widely observed in complex networks. This framework makes the widely applied small world property statistically more rigorous.
(7) Statistical model, estimation, and hypothesis testing methods for community structure in a population of networks for analyzing data from multi-subject fMRI neuroimaging studies. These methods enable neuroscientists to assess differences in functional organization of brain networks between a population of patients and healthy controls.
(8) Methods for estimating the latent positions of a network generated from the hyperbolic latent space model with community structure.
(9) A generative model for correlation networks, known as the correlation generalized exponential random graph model, which characterizes the topological structure of a correlation network using an exponential model on summary statistics of the observed graph.
(10) A network monitoring technique which combines the use of random graph modeling and statistical process monitoring (SPM) techniques.
The overarching aim of this project was to develop dynamic network models and algorithms for representing, analyzing, monitoring, and predicting complex interactions from spatiotemporal data. These developments together represent important contributions in many statistical problems on analysis of spatiotemporal data. We have also engaged actively in collaborating with researchers from Psychology, Education, Neuroscience, and Microbiology. We have applied the models and methods developed in this project for various application problems from those fields.
The project has supported the research of 4 PhD students in Statistics and 1 PhD student in Psychology. In addition, 1 MS student and 1 Undergraduate student were involved in the project. Three of the PhD students were partially supported by funds from this project. The PI has taught a graduate special topics course on statistical inference in networks in Spring 2020 that has incorporated knowledge gained from this project.
The project team has disseminated the results to the communities of interest through invited talks at several conferences including the 2018 and 2020 CMStatistics conferences, the 2021 Network conference, the 2019 IISA conference, 2019 IEEE Data Science workshop, and the 2019 ABS conference.
The PI has organized sessions at the 2020 CMStatistics conference, and the 2019 IEEE Data Science Workshop in Minneapolis, MN. The Project team has also attended the ATD conferences in 2018, 2019, and 2020. The PI has presented in the special topic contributed session organized on behalf of the ATD program at the 2021 JSM conference.
The results of the research from this project has been published in several statistical methodology journals including the Annals of Applied Statistics, Annals of Statistics, Sankhya, machine learning conferences, including NeurIPS, as well as several interdisciplinary journals. A large number of articles with the results are also posted in the preprint server Arxiv.
Last Modified: 04/30/2022
Modified by: Subhadeep Paul
Please report errors in award information by writing to: awardsearch@nsf.gov.
