Combing through news articles could help us predict food crisis outbreaks almost a year in advance!

One of the biggest detriments of the ongoing Russia-Ukraine conflict was the treacherous manner in which global food security was vandalised in many different parts of the globe. The war brought the Horn of Africa — a region already struggling with malnourishment and food scarcity — further to its knees, and its effects reverberated through existing food systems worldwide.

While there are ways in which we can predict an incoming food crisis, the prediction systems are too slow most of the time. In such large-scale disasters, preparedness is vital. Therefore, we desperately need to think outside the box to maximise the time to respond to such problems. Luckily, the platform you're reading this article on might be the very solution to this issue.

A group of researchers have developed a machine-learning model that scours through news articles to pick up on key food crisis-related terms and phrases and uses that information to predict which areas face an elevated risk of food insecurity. According to the researchers, this is a vast improvement over current methods.

"Our approach could drastically improve the prediction of food crisis outbreaks up to 12 months ahead of time using both real-time news streams and a predictive model that is simple to interpret," says Samuel Fraiberger, one of the study's authors.

The researchers note that news coverage of risk factors triggering a food crisis often ramps up significantly before the devastating event. These "real-time, on-the-ground accounts" of local developments can be quantified and used as an early-warning system for impending food crises in the area.

To validate that news can be relied on for such predictions, the researchers combed through 11 million articles focusing on nearly 40 food-insecure countries published between 1980 and 2020. Doing so, they managed to extract food-crisis-related phrases from these. Comparing this data to historical food-insecurity risk factors such as conflict fatality counts, rainfall, and food price changes, they found a high correlation between food crisis coverage and real-life event occurrence, indicating that news stories can accurately predict such conditions.

As mentioned earlier, these news-based predictions were quick and highly localised, beating out traditional methods of traditional measurements. There was also some merit in supplementing traditional predictive measures with this news-coverage technique, suggesting a "hybrid" model might be the best way to go.

In 2021, almost a billion people suffered from hunger, with the COVID-19 pandemic exacerbating conditions worldwide. In order to make sure we reach our second sustainable development goal of zero hunger, such predictive measures will come in very handy.

Recommended stories: