In this project, we developed a model of Ebola spread by using innovative big data analytics techniques and tools. We used massive amounts of data from various sources including Twitter feeds, Facebook and Google. This data is then fed into a decision support system that models the spread pattern of the Ebola virus and creates dynamic graphs and predictive diffusion models on the outcome and impact on either a specific person or a specific community. As a result of this research, computational spread models for Ebola in the U.S. are created, potentially leading to more precise forward predictions of the disease propagation and tools to help identify individuals who are possibly infected, and perform trace-back analysis to locate the possible source of infection for a particular social group.  Besides collaborating with FIU, we also closely collaborated with LexisNexis (LN), which is a leading big data company and a member of our I/UCRC for Advanced Knowledge Enablement. LexisNexis has provided the large amount of data about relationship of the people in US, and we combined it with data analytics techniques and tools to model disease spread patterns. In this part of research we used Cloud Computer system located in our College at FAU as well as LN’s High-Performance Computer Cluster (HPCC), which is intended for big data applications. We performed modeling, analytics, and development of a Decision Support System (DSS), which provides a probabilistic outcome of Ebola impact on either a specific person or a community at a specific location. This information is then fed to FIU’s Terafly system for geospatial mapping and other services.

Jointly with the LN research team, we created people clusters based on proximity and built a model using weighted scores, which approximate physical contacts. In creating people clusters, we used public record graph to calculate distances between an affected person and his/her relatives and friends. Base on this model, we developed disease propagation path.

This work represents an improvement over previous state-of-the-art, because we used innovative data analytics techniques and the latest HPCC technology to developed models of Ebola spread. Mathematical compartmental models have been applied to predict the behavior of disease outbreaks in many studies. These models aim to understand the dynamics of a disease propagation process and focus on partitioning the population into several health states. With information from multiple sources indicating infected individuals and their personal relationships and social groups, dynamic graphs can be created, and predictive diffusion models can be used to study key issues of Ebola epidemics, e.g., location, time and number of expected new cases. Two fundamental diffusion models are Independent Cascade Model (IC) and Linear Threshold Model (LT), both of which follow an iterative diffusion process where infected nodes infect their uninfected neighbors with certain probabilities. Based on fundamental models, we developed advanced propagation models to estimate an influence function by examining past and newly infected notes and predict subsequent infections. Our program is developed to identify and visualize families and tightly connected social groups who have had some contact with Ebola patients. Tracking and containing this disease requires enormous resources. Our system provides a proactive approach to reasonably reduce the risk of exposure of Ebola spread within a community or a geographic location.

There are several end-user products based on our research. The main objective is to feed Ebola spread patterns, which we obtained from our models, into Decision Support System (DSS). The inputs into DSS are also information about social groups or individual persons. Social groups can be for example nurses and doctors who had contacts with an Ebola patient, or passengers who travelled in the same plane with an Ebola patients, or family members of an Ebola patent, and others. Based on spread patterns, the DSS then calculates probabilities for a social group or a given person to get affected with Ebola.

The mobile interface will further allow the people to enter the signs of Ebola such as ‘Fever’ specifically above 100.4^o, chills, headache and vomiting, myalgia, intense weakness. While these signs are very common with other diseases as well such as Malaria and Typhoid, therefore laying over these impacts on the geo-coordinates of having the person in a geographic proximity of a person of a community with impact of Ebola will provide more focused results with higher accuracy.

The economic impacts can be tremendous by predicting outbreaks of the deadly Ebola virus (or any other epidemics) and directing potential victims of disease to the nearest suitable medical facility. This research can also help in indicating the areas in which new facilities should be opened, where disease outbreaks are beginning to occur, and how they are likely to expend.

Last Modified: 11/02/2016
Modified by: Borko Furht