Scientists create tool that can predict food insecurity

Researchers have created a tool they claim can be used to forecast the prevalence of individuals within a country that may have insufficient access to food.

Elisa Omodei, an author of the study and Assistant Professor at the Department of Network and Data Science at the Central European University, said that the tool can predict food insecurity up to 30 days in advance and it could aid decision makers in countries at risk of food insecurity by helping to facilitate more timely responses.

Using food consumption data from Burkina Faso, Cameroon, Mali, Nigeria, Syria and Yemen (countries that have recently experienced food insecurity), the researchers analysed data from 2018-2022.

The authors then enhanced their tool with data on conflict-related fatalities, food prices, extreme weather events and the occurrence of Ramadan throughout this period. The tool was then used to estimate the prevalence of households at risk of insufficient access to food between October 2021 and February 2022.

From their research, the team claim that the tool was able to forecast the prevalence of food insecurity within Yemen and Syria with 99 percent accuracy one day into the future and with 72 percent and 47 percent accuracy respectively 30 days into the future.

However, the researchers admitted that when it came to data for the four remaining countries, the tool’s predictions were not as accurate. The researchers said that this was because there was less available food consumption data compared to Syria and Yemen.

Will the invasion of Ukraine spark a global food insecurity crisis?

“This highlights that the tool’s forecasts are more accurate when using food consumption data collected at regular intervals over long periods of time and across a broad range of geographic areas,” said the researchers.

Explaining how the tool could be useful, the authors suggest that their tool could complement existing techniques for modelling food insecurity by providing rapidly available forecasts using real-time data.

Going forward, the authors say that the tool’s predictions could be improved by incorporating mobile phone data or automated text mining of news.