Infectious Diseases Outbreak Prediction using Geolocation Data with Machine Learning

Although the recent COVID-19 pandemic is arguably the most disastrous pandemic of the 21st century, disease outbreaks regularly occur, albeit on a geographically limited scope. For example, in the past decade alone, previously unknown diseases such as the Severe Acute Respiratory Syndrome Coronavirus (SARS-CoV-1), the Middle East Respiratory Syndrome Coronavirus (MERS), and the Influenza A virus subtype H7N9 inflicted havoc upon the communities and regions where they emerged. Likewise, humanity is not safe with known communicable diseases such as Ebola, Zika virus, cholera, tuberculosis, and measles. For the least developed countries, outbreaks of communicable diseases poses particular challenges. Indeed, low-income countries lack adequate sanitation and hygiene infrastructure and have limited financial resources to trace, identify, test, and quarantine suspect cases. Additionally, unexpected disease outbreaks put pressure on the already overburdened and underfunded healthcare.

While the World Health Organization (WHO) advocates for methods to rapidly pinpoint and isolate infected people before there is a major outbreak, there exist no pragmatic methods to detect and prevent infectious diseases before they spread. The few promising solutions expect the availability of high-level infrastructure, require advanced know-how, and necessitate significant financial resources; thus, are not practical in the context of developing countries. When everyone needs better information, including epidemic disease researchers, government authorities, and international organizations, digital information, and surveillance technologies are relevant tools to collect data and reliable evidence to support public health decision-making.

This project proposes to leverage the ubiquitous mobile payment systems in order to collect location data to be used to detect infectious diseases in their early stage and to help decisionmakers take adequate countermeasures to prevent their spread. The proposed approach hinges on the possibility of estimating the geographic location of each mobile payment transaction and the identification of the customer. Such data, together with live data of infection test results, allows the development of models that enable the detection of infectious diseases before they spread widely and can be used to identify clusters of infected people and estimate the disease’s super-spreading potential. The proposed approach is low-cost, pragmatic, and is applicable to the context of developing countries. The present project involves the collaboration between the University of Rwanda, Universite Gaston Berger (Senegal), and LocationMind Inc (Japan).