Using Predictive Analytics for Humanitarian Goals

Using Predictive Analytics for Humanitarian Goals - Research Exchange Lab
Members of the UVA Humanitarian Collaborative’s Predictive Analytics team meet with Save the Children in London to work on their modeling project.

As the humanitarian operational community invests in data-driven strategic planning and programming, they’re developing the forecasting tools that will help predict the timing, magnitude, and duration of mass-displacement and crisis events that affect children and their families. This theme brings together faculty from Batten, the Departments of Statistics, Politics, and Systems Engineering, along with the School of Data Science, among others to build and evaluate useful displacement models that can guide humanitarian response.

The Team

Faculty, Staff and Students

Our core faculty and staff oversee the Global Policy Center and promote research and education within the humanitarian sector.

Development of a Broad Predictive Analytics Platform for Humanitarian Crises

In partnership with SCI, UVA faculty and students are expanding on the SCI displacement forecasting prototype to create a broader platform. The team is:

  • identifying knowledge gaps in the modeling of displacement events;
  • engaging in data collection, cleaning, and collating of data related to understanding displacement events;
  • identifying both structural and dynamic variables that influence the occurrence, duration, and magnitude of displacement events;
  • developing a statistical model the forecast future displacement events at reasonable intervals;
  • demonstrating the validity of the model using historical qualitative and quantitative data, statistical cross validation, and “ground-truthing” working directly with Save the Children staff working on children migration issues worldwide; and
  • working together to begin the process of developing a user-interface so that relevant actors can generate forecasts based on needs of their constituencies.

A validated tool for real-time predictive analytics is of use across the humanitarian response landscape. The development of such a tool will generate knowledge spillovers in applied statistics and forecasting as well as for forecasting efforts in related policy arenas.

Network-based Mobility Modeling for Complex Humanitarian Emergencies

Human mobility drives both the spread and impact of infectious diseases. Complex humanitarian crises generated by civil conflict or natural disasters frequently include some combination of mass displacements, damage to infrastructure, lack of adequate food and shelter, and disease outbreaks. This combination of factors often results in the spread of disease as those displaced by the crisis seek temporary or permanent refuge in other locations and come into contact with other populations. Identifying the effect of a policy response or intervention is challenging. Likewise, current outbreak response models are unable to incorporate the social impact of the outbreak itself, let alone other factors, in estimating the spatial spread of the outbreak.

In partnership with the Global Infectious Disease Institute, UVA Faculty and students will address these gaps and provide a means to use quantitative data-driven methods to inform outbreak responses and the effectiveness of policies in the midst of complex humanitarian emergencies. This project develops a novel simulation framework to bring together both migration attractors as well as repellers while simulating the spread of disease. It leverages previous work on epidemic modeling, network construction, and knowledge and quantitative models for estimating patterns of global migrant flows.

The project advances core knowledge regarding the quantitative treatment of complex humanitarian emergencies that feature infectious disease outbreaks and provide better risk assessment and response planning.