Tuesday, June 02, 2020

Location-Based Social Simulation for Prescriptive Analytics of Disease Spread

Building upon our previous work on Location-Based Social Networks (LBSNs) and how agent-based modeling could provide an alternative to real world data sets, in the latest SIGSPATIAL Special Newsletter, we (Joon-Seok Kim, Hamdi Kavak, Chris Rouly, Hyunjee Jin, Dieter Pfoser, Carola Wenk, Andreas Zufle and myself) have an article entitled "Location-Based Social Simulation for Prescriptive Analytics of Disease Spread."

In this article we discuss a geographically explicit agent-based model that we have been developing that is capable not only of simulating human behavior but also able to create synthetic but realistic LBSN data based on human patterns-of-life. Furthermore, in the article we discuss how such data and models can be used to explore the parameter space of possible prescriptions to find optimal strategies (or policies) to achieve a desired system state and outcome. We refer to such a search for optimal policies as prescriptive analytics. (for readers wishing to learn more about prescriptive analytics please see the 1st ACM KDD Workshop on Prescriptive Analytics for the Physical World).

To give an example of such prescriptions, in the article we make use of a simple hypothetical disease model and explore two prescribed policies to mitigate the spread of the disease. The first policy requires all agents to wear simulated Personal Protective Equipment (PPE) that reduce the chance of infection by 50%. The second policy enforces strict social distancing measures onto a fixed proportion of 50% of the population. Those who follow the social distancing order avoid recreational site visits from meeting people although they still go to restaurants. In addition to these two policies, as a baseline, we also ran a “null-prescription” in which no intervention was prescribed. We find that the social distancing prescription was extremely effective. On the other hand, our simulation results for PPE policy showed that merely wearing protective gear without any change in behavior has no significant effect (for the case of this disease).

If this type of research is of interest to you, below we provide the abstract to the paper, a movie of a representative simulation run, some of our results of the prescriptions described above and a link to the paper itself. Further information about the model and data can be found at https://geosocial.joonseok.org/p/epidemic.html and the data is available at https://osf.io/e24th/. Also as we are currently going through COVID-19, we thought a a brief write up and links to some disease models and discussions of modeling efforts related to it was also appropriate to include.

Human mobility and social networks have received considerable attention from researchers in recent years. What has been sorely missing is a comprehensive data set that not only addresses geometric movement patterns derived from trajectories, but also provides social networks and causal links as to why movement happens in the first place. To some extent, this challenge is addressed by studying location-based social networks (LBSNs). However, the scope of real-world LBSN data sets is constrained by privacy concerns, a lack of authoritative ground-truth, their sparsity, and small size. To overcome these issues we have infused a novel geographically explicit agent-based simulation framework to simulate human behavior and to create synthetic but realistic LBSN data based on human patterns-of-life (i.e., a geo-social simulation). Such data not only captures the location of users over time, but also their motivation, and interactions via temporal social networks. We have open sourced our framework and released a set of large data sets for the SIGSPATIAL community. In order to showcase the versatility of our simulation framework, we added disease a model that simulates an outbreak and allows us to test different policy measures such as implementing mandatory mask use and various social distancing measures. The produced data sets are massive and allow us to capture 100% of the (simulated) population over time without any data uncertainty, privacy-related concerns, or incompleteness. It allows researchers to see the (simulated) world through the lens of an omniscient entity having perfect data.

Screenshot of the epidemic simulator depicting the French Quarter, New Orleans, LA, USA.

New cases and SEIR epidemic course.

Full Reference:
Kim, J-S., Kavak, H., Rouly, C.O., Jin, H., Crooks, A.T., Pfoser, D., Wenk, C. and Zufle, A. (2020), Location-Based Social Simulation for Prescriptive Analytics of Disease Spread, SIGSPATIAL Special, 12(1): 53-61. (pdf)

The Washington Post's Disease Model
While this post is not about COVID per se, if you are interested in disease models the Washington Post had a great article about COVID several months ago entitled "Why outbreaks like corona virus spread exponentially, and how to “flatten the curve”." This article generated a lot of discussion such as on the SIMSOC Mailing list and was citied in a paper in the Journal of Artificial Societies and Social Simulation (JASSS) entitled  "Computational Models That Matter During a Global Pandemic Outbreak: A Call to Action." Other goods discussions on COVID related models (particularly agent-based models) can be found on Review of Artificial Societies and Social Simulation (RofASSS) website (here), the CoMSES Net Discourse Forum (along with links to past epidemic models) and the Sociology and Complexity Science Blog has some very good posts on modeling and public health.

Tuesday, May 26, 2020

Crowdsourcing Street View Imagery: A Comparison of Mapillary and OpenStreetCam

In the past we have written extensively on Volunteered Geographic Information (VGI) such as OpenStreetMap or Twitter. However, we have not really explored Street View Imagery  (SVI), well not until now. Within the realm of VGI, SVI has emerged in recent years as a novel and rich source of data on cities from which geographic information can be derived.

Perhaps the most well-known example of SVI utilization is that of Google Street View (GSV). While SVI has been traditionally collected by governmental agencies and companies alike, we are now also witnessing the emergence of Volunteered Street View Imagery (VSVI), which relies on a crowdsourced effort to provide geotagged street-level imagery coverage of traversable pathways (e.g., a street or trail). Such imagery, similar to GSV, provides detailed information about the location of objects such as cars, road markings, traffic lights and signs, and allows for the automatic extraction of features at scale. Such imagery can also be mined using machine learning algorithms to automatically derive points of interest (POI) databases (e.g., locations of coffee shops and fire hydrants) without the intervention of the citizen.

To explore VSVI we have just published a new paper entitled: "Crowdsourcing Street View Imagery: A Comparison of Mapillary and OpenStreetCam" in the ISPRS International Journal of Geo-Information. In this paper we examine VSVI data collected from two different platforms: Mapillary and OpenStreetCam (OSC) for four metropoiltan areas in the United States (i.e., Washington (District of Columbia), San Francisco (California), Phoenix (Arizona), and Detroit (Michigan)). Both of these online platforms accept sequences of images captured from mobile devices and uploaded via an app on the device (like those shown in the image to the right). Images are geolocated using the device’s global positioning system (GPS). More specifically the paper examines:
  • the level of spatial coverage of each platform in order to assess the overall potential of such platforms to provide adequate coverage of geographic information.
  • user contribution patterns in Mapillary and OSC in order to understand how users are contributing to these platforms.
Results from our systematic and quantitative analysis of these two emerging VGI sources indicate that most Mapillary and OSC contributions occurred along control-access highways and local roads, and that the overall coverage in these sources is variable in comparison to an authoritative source (i.e., TIGER). Furthermore, our results showed that while the number of contributors varied across sites, only a few contributors were responsible for producing most of the raw data. User contribution patterns were also different in Mapillary and OSC. Specifically, we found that while patterns in coverage were variable for the different OSC sites, coverage patterns in Mapillary tended to be similar among sites. This finding may be linked to several factors, including differences in mapping practice, or issues with participation inequality, a topic that has been highly researched for other VGI platforms such as OSM, but which is still lacking within VSVI. Lastly, user contributions in Mapillary tended to be higher around 8:00 am, 1:00 pm and 5:00 pm (local time). This finding suggests that VSVI contributions tend to coincide with the morning and afternoon commute, and the lunch hour of the contributors.

If you wish to find out more about this work below we provide the abstract to the paper, a visual flowchart of our workflow and some of our our results. The full reference and link to the paper is provided at the bottom of the post.

Over the last decade, Volunteered Geographic Information (VGI) has emerged as a viable source of information on cities. During this time, the nature of VGI has been evolving, with new types and sources of data continually being added. In light of this trend, this paper explores one such type of VGI data: Volunteered Street View Imagery (VSVI). Two VSVI sources, Mapillary and OpenStreetCam, were extracted and analyzed to study road coverage and contribution patterns for four US metropolitan areas. Results show that coverage patterns vary across sites, with most contributions occurring along local roads and in populated areas. We also found that a few users contributed most of the data. Moreover, the results suggest that most data are being collected during three distinct times of day (i.e., morning, lunch and late afternoon). The paper concludes with a discussion that while VSVI data is still relatively new, it has the potential to be a rich source of spatial and temporal information for monitoring cities.

Keywords: Crowdsourcing; Volunteered Geographic Information; Street View Imagery; Mapillary, OpenStreetCam
Overview of methodology

Spatial distribution of road networks.
Spatial comparison of roads in kilometers.

Full Reference: 
Mahabir, R., Schuchard, R., Crooks, A.T., Croitoru, A. and Stefanidis, A. (2020), Crowdsourcing Street View Imagery: A Comparison of Mapillary and OpenStreetCam, ISPRS International Journal of Geo-Information. 9(6), 341; https://doi.org/10.3390/ijgi9060341 (pdf)

Thursday, May 21, 2020

A Semester of Spatial Agent-based Models

So draws an end of another semester and as it is becoming a bit of tradition, here is a post highlighting some of the class projects from my graduate class  entitled "Spatial Agent-based Models of Human-Environment Interactions". As with many of my courses, students were expected to complete a end of semester project, in this case, develop an agent-based model that explores some aspect of related to the course theme of human-environment interactions.  

For several of the students this was their first exposure to either agent-based modeling or utilizing geographical information in the modeling process. In the movie below a selection of these projects can be seen. The projects ranged from exploring how farming practices impact erosion, water reuse practices within agriculture, to the spread of diseases, deciding to evacuate during a disaster, to that of war gaming or how  zooplankton impacts Basking Shark shoaling behavior. As can be seen the movie, the models ranged from abstract spatial representations to those utilizing geographical information as a foundation of their artificial worlds. Many of the models where created using NetLogo (including one using LevelSpace) while others chose to utilize MASON or Mesa.

I would like to thank the Students of CSS 645: Spatial Agent-based Models of Human-Environment Interactions for their participation both in person and virtually in the class.  

Sunday, May 03, 2020

Utilizing Agents To Explore Urban Shrinkage

While more people are living in urban areas than ever before, and this is expected to grow in the coming decades, this growth is not equal. Some cities are actually shrinking, such as Detroit in the United States. The causes of urban shrinkage have been the source of much debate but can be broadly attributed to a combination of factors relating to deindustrialization, suburbanization (i.e., urban sprawl), and demographic withdrawal. The result of shrinking cities, especially in and around the traditional downtown core of the city results in many problems, such as population loss, economic depression (due to loss in tax revenue), a growth in vacant properties, and the contraction of the land and housing markets.

To explore this phenomena, at the upcoming 2020 Spring Simulation Conference we have a paper entitled "Utilizing Agents To Explore Urban Shrinkage: A Case Study Of Detroit." The motivation for this paper is to explore the housing market in a shrinking city from the micro-level, specifically based on individuals trading interactions via an agent-based model stylized on spatially explicit data of Detroit Tri-county area. Our agent-based model demonstrates the potential of simulation to explore urban shrinkage and potentially offers a means to test polices to alleviate this issue. For readers wishing to know more about this work, below we provide the abstract to the paper, some figures sketching out some of model logic,  a sample of results and a movie of a representative model run. Similar to our other works, we have a more detailed description of the model following the Overview, Design concepts, and Details (ODD) protocol along with the source code and data needed to run the model at: http://bit.ly/UrbanShrinkage. We do this to aid replication and for others to extend if they see fit. As normal, any thoughts or comments are most welcome.

While the world’s total urban population continues to grow, this growth is not equal. Some cities are declining, resulting in urban shrinkage which is now a global phenomenon. Many problems emerge due to urban shrinkage including population loss, economic depression, vacant properties and the contraction of housing markets. To explore this issue, this paper presents an agent-based model stylized on spatially explicit data of Detroit Tri-county area, an area witnessing urban shrinkage. Specifically, the model examines how micro-level housing trades impact urban shrinkage by capturing interactions between sellers and buyers within different sub-housing markets. The stylized model results highlight not only how we can simulate housing transactions but the aggregate market conditions relating to urban shrinkage (i.e., the contraction of housing markets). To this end, the paper demonstrates the potential of simulation to explore urban shrinkage and potentially offers a means to test polices to alleviate this issue.

Keywords: Urban Shrinkage, Housing Markets, Detroit, Agent-based Modeling, GIS 

Agents Decision Making Process.

The sequences of all function events in the model are displayed by this UML diagram, which demonstrates the model flow, dynamic and interaction among the different components of the model.

Average Results where: (a) demand exceeds supply; (b) equal demand and supply; (c) supply exceeds demand for each different housing sub market.

Jiang, N. and Crooks, A.T. (2020), Utilizing Agents to Explore Urban Shrinkage: A Case Study of Detroit, 2020 Spring Simulation Conference (SpringSim’20), Fairfax, VA. (pdf)

Thursday, April 30, 2020

Exploring the Effects of Link Recommendations on Social Networks

Most people today are actively engaged on at least one social networking site, enabling individuals to keep in touch with old friends, connect with new people, and rapidly disseminate information to all. The method by which users find and link up with others online is often assisted by recommendation systems. A common technique utilized by online social networking sites (e.g., LinkedIn, Facebook) is to make link recommendations based upon friends of friends, or shared mutual connections. This method exploits a user’s social network structure, and specifically transitivity, to predict that a user will be interested in connecting with an individual who is also connected with that user’s friends, (i.e., “I am more likely to like someone who several of my friends like, than someone chosen at random”).

Despite the wide use of recommendation algorithms, little is known about the way in which recommendation systems impact the structure of online social networks. To address this problem at the upcoming (now virtual) 2020 Spring Simulation Conference we have a paper entitled "Exploring the Effects of Link Recommendations on Social Networks: An Agent-Based Modeling Approach."

This paper contributes to this limited area of (publicly available) research by demonstrating how a stylized agent-based model can be used to explore societal, network-level effects of commonly used online link recommendations from the bottom up. Below we provide the abstract to the paper, the steps the model takes to generate the online social network, the types of metrics outputted by the model and a selection of some of the results. While at the bottom of the post we provide the full reference to the paper. Further details about the model, in the Overview, Design Concepts, and Details (ODD) format along with the Python source code can be found at https://www.comses.net/codebase-release/3203a44a-fcb0-4957-a3b7-8323f829c0c4/

The vast majority of recommender system research has focused on improving performance accuracy, while limited work has explored their societal, network level effects. This paper demonstrates how simulation can be used to investigate macro level effects of online social network link recommendations, such as whether these technologies may be fragmenting or bridging communities of individuals. An agent-based model is presented that generates stylized online social networks with different percentages of real world contacts and link recommendations. Results show that networks with higher percentages of recommendation-based links produce more clustered, distinct, and dispersed communities, suggesting that these technologies could fragment society. Furthermore, scale-free network properties diminished with higher percentages of recommendations, suggesting that these technologies could be contributing to recent findings that social networks are at most ‘weakly’ scale-free. Building upon this research, further simulation work could inform the design of link recommendation algorithms that help connect both individuals and communities.

Keywords: online social network, social network analysis, mutual connection link recommendation system, friend-of-friend recommender, agent-based modeling.
Online network generation process
Social Network Analysis definitions for metrics output by this model
The effect of link recommendations on mean clustering coefficient and modularity. Error bars represent one standard deviation.
Probability density functions showing the degree distribution of Online networks (beginning top left and increasing left to right, top to bottom) with link recommendation percentage levels: 0, 10, 20, 30, 40, and 50. The blue line represents the empirical data, and red and green dotted lines represent fit lines corresponding with the power-law and lognormal distributions, respectively.

Full Reference: 
Sibley, C. and Crooks, A.T. (2020), Exploring the Effects of Link Recommendations on Social Networks: An Agent-Based Modeling Approach, Spring Simulation Conference (SpringSim’20), Fairfax, VA. (pdf)

Update: Our paper was selected as runner-up for best paper.