Building an infrastructure damage index to predict human displacement due to conflict or natural disasters:

Save the Children’s Migration and Displacement Initiative (MDI) was established in 2016 to respond to the increasing impact of displacement and migration crises upon children. Four times as many people are displaced today than 15 years ago. Displacements are also lasting longer, with the average duration of a displacement now 10 years. Children are disproportionately affected, with severe impact on their health, education and wellbeing. 

MDI’s Predictive Displacement Project is currently developing a predictive analytics model for anticipating the future scale and duration of conflict-driven displacement crises. Without this data, humanitarian actors struggle to plan interventions that meet the needs of the displaced population. Lack of demographical data also limits how effectively they can meet the specific needs of vulnerable groups within displaced populations, especially children. To resolve this, MDI is working to predict these characteristics of displacements, and so dramatically improve the quality of humanitarian responses. 

Early prototyping suggested a strong link between levels of infrastructure damage and the scale and length of displacements, but there is no reliable or systematically available dataset available that describes this factor. The goal of this hackathon is to explore and prototype different options for generating infrastructure damage values that can be applied to MDI’s predictive analytics model. 

This hackathon is open to all UVA students. Students will work in teams over a two week period to generate methodologies that can accurately measure the extent of infrastructure damage in a region. The primary data sources will be geospatial and land-satellite images, as well as text from wire news services, reports, and social media. Participants can use any methods they want to generate these measurements. We will be providing support to students who would like to use this opportunity to learn new skills. We will be asking each group to organize and document their work in a GitHub repository so that the code can be easily shared and incorporated into students’ professional portfolios.

The event will be taking place over two weeks with two main events occurring live over Zoom. On Saturday, October 10, we will be holding an introductory event in which we will describe the Predictive Displacement Project in detail, along with the data sources we will be using for this hackathon. We will provide examples of similar work to help participants brainstorm their own approach to the problem. We will also make sure all of the participants have access to our Slack workspace, computing resources, and our GitHub organization. 

Over the next two weeks, participants will work with their teams to generate a methodology that measures the extent of infrastructure damage. We will be providing support, mentorship, and advice to participants during this time.

The final event will occur on Saturday, October 24. Every group will briefly present their work at that time.   

All teams will receive feedback on their submission, though the event is non-competitive and will not be judged as such. However, all participants will receive:

  • A potential opportunity to further collaborate with Save the Children through their academic term and beyond
  • An opportunity to learn and experiment with cutting edge technologies, e.g. natural language processing, imagery analysis.
  • A chance to make a genuine positive social impact on humanitarian aid to displaced people. 
  • Interaction with data science, machine learning, public policy, and humanitarian experts.
  • A strong professional portfolio piece
  • A certificate of Appreciation from Save the Children, and coverage in a Save the Children's media content.

 Questions? Jon Kropko (jkropko@virginia.edu)                                       

                                  Register


                     This event is done in partnership with

School of Data Science Logo