Request support on coordination, information management, integration for nutrition outcomes or technical nutrition in emergencies assistance.

Request Support
Publication: 18.02.2021

Nutrition Machine Learning Workshop


The objective of this workshop is to define the GNC approach on Predictive Analytics for the forecasting of Nutrition outcomes. The GNC, under the auspicious of the GNC Technical Alliance, is embarking on a new initiative on Nutrition Predictions. When exploring the existing work in this area, there are several initiatives and projects looking at projections in general and nutrition predictions in particular (see agenda sessions below).

In order to avoid duplication of efforts, maximize learning, and decide on the best way forward with nutrition projections, two half-day workshops are organized to learn about the pros and cons of different approaches to projections that have been tried so far, and what would be needed to develop a forecasting model that is useful for all countries and decision-makers for Nutrition.

Day 1 Agenda. What is out there?
Facilitators: Hassan Ali Ahmed & Anna Ziolkovska

Session title

Presenter (s)


Introduction of the participants and the purpose of the workshop

Anna Ziolkovska

The purpose of the session is that we are all clear on what we need to achieve and how we get there

IPC Acute Malnutrition forecasting

Douglas Jayasekaran

This is methodology currently used by most of the countries (does not involve AI or any mathematical modeling; mostly trend analysis with assumptions)

Geospatial data solutions for humanitarian action

Margaretha Barkhof

The purpose of this session is to introduce the work developing global level resources for humanitarian action (i.e., the humanitarian elements of the UNICEF GeoHub)


Nicholas Haan


The IPC Secretariat has established the Advanced Technology and Artificial Intelligence (ATARI) working group to improve forecasting for the IPC FS analysis. This presentation is to share experiences that can be potentially transferred for enhancing AM predictions.

Hunger map

Jonathan Rivers, Elisa Omodei

The purpose of this session is to present the predictive model that WFP developed to nowcast the prevalence of people with insufficient food consumption (

Modelling Early Risk Indicators to Anticipate Malnutrition (MERIAM)

Alice Stevenson, Ellyn Yakowenko, Pascal Debons

Modelling Early Risk Indicators to Anticipate Malnutrition (MERIAM), is a four-year research project funded by the UK government, which seeks to identify, test, and scale up cost-effective means to improve the prediction and monitoring of acute malnutrition, through the use of open access secondary data.

South Sudan forecasting

Saeed Rahman, Qutab Alam, Ismail Kassim

Used predictive modeling techniques to estimate the burden of acute malnutrition and severity mapping in the absence of recent anthropometric data

Summary of the day and next steps



Day 2 Agenda. What is out there? + Reflections on the way forward
Facilitators: Louise Mwigiri & Douglas Jayasekaran 

Session title

Presenter (s)


Summary of the Day 1 discussions and introduction to the day 2

Chairs of day 1


Overview on Obesity/overweight surveillance systems. Use case: Scotland

Alex Hutchison


This is an overview of the project: “Analyses of the “Growing up in Scotland (GUS)” child cohort to inform the design of obesity/overweight surveillance systems internationally”.

More information about the Data for Children Collaborative with UNICEF is here:

Prediction of acute malnutrition prevalence in three countries in Eastern Africa

Mara Nyawo

UNICEF Eastern and Southern Africa Regional Office has been working with LSHTM to develop a model that will predict GAM and SAM for small geographical areas. This presentation will cover the concept of the research.

Future directions of the IPC

Francesco Mosconi

IPC plans in machine learning and future directions

Initial GNC project approach (pre-workshop)

Anna Ziolkovska

Preliminary GNC NIS WG thoughts on the forecasting

Facilitated discussion to clearly define the forecasting problem GNC wants to solve


Some questions to think about:

  • Does the presented approach make sense?
  • Is it feasible to select as an outcome acute malnutrition?
  • Would it be possible to define the outcome as the regression problem (continuous data)?
  • What accuracy and lead time we can realistically achieve?
  • At what admin level can we predict the outcome?

Facilitated discussion on how GNC should approach forecasting for Nutrition outcomes


Clear way forward agreed on what GNC should do to improve its forecasting.

Summary and next steps

Chairs of day 2

Summary of the main agreements during the workshop


to receive GNC newsletters