Nowcasting food insecurity on a global scale

Abstract

Lack of regular physical or economic access to safe, nutritious and sufficient food is a critical issue affecting millions of people world-wide. Estimating how many and where these people are is of fundamental importance for governments and humanitarian organizations to take informed and timely decisions on relevant policies and programmes. In this study, we propose a machine learning approach to predict the prevalence of people with insufficient food consumption and of people using crisis or above crisis food-based coping when primary data is not available. Making use of a unique global data set, we show that the proposed models can explain up to 78% of the variation in insufficient food consumption and crisis or above food-based coping levels. We also show that the proposed models can be used to nowcast the food security situation in near-real time and propose a method to identify which variables are driving the changes observed in predicted trends, which is key to make predictions serviceable to decision makers.

On the forecastability of food insecurity

“…In this context, our study represents an initial step towards the application of forecasting approaches to food insecurity at a high spatial and temporal granularity. Our results confirm that nowcasting or one-step-ahead forecasting are feasible, as reported in recent studies [36, 38], but long-term forecasts are challenging and strongly conditioned by data availability. The methods presented in this study come with limitations, and they could be further improved through several approaches.…”

“…Okori and collaborators first proposed to use machine learning models to predict whether a household is in famine or not from household socioeconomic and agricultural production characteristics [33, 34]. The effort of predicting levels of insufficient food consumption has been tackled in the context of Malawi, in a study where the authors built a model trained on 2011 data to estimate the situation in 2013 [35], and more recently in a work proposing a model to nowcast sub-national levels of insufficient food consumption on a global scale [36]. Both studies propose methods to predict the current situation when primary data is not available, but they do not address the challenge of making projections for the future.…”

Explaining food security warning signals with YouTube transcriptions and local news articles

“…Some examples may be spatial information (e.g., population density, land use, soil quality), volunteered geographical information (number of hospitals and schools, number and details about violent events) and economic indicators (e.g., price of representing goods). Recent literature has shown how these heterogeneous data can be exploited to predict food security indicators through advanced data science methods [4,11,21], e.g., multi-branch neural networks able to integrate data of different types and at different scales, by also taking into account spatial and temporal context. Nevertheless, while the performance of these approaches seem to be promising, they are still far from being optimal, and strongly dependent from the study area taken into account (e.g., availability and quality of the data may not be the same over different country, as well as the correlation with food security indicators).…”

Machine Learning Guided Outlook of Global Food Insecurity Consistent with Macroeconomic Forecasts

“…For example, Mwebaze et al (2010); Okori and Obua (2011) predict household famine in Uganda between 2004, Vu et al (2022 predict changes in household level food insecurity in Vietnam, and Lentz et al (2019) predict food insecurity for village clusters in Malawi. Martini et al (2021); Foini et al (2022) develop high-frequency now-casting capabilities at a global scale focused on the food consumption score and Balashankar et al (2021) extract leading signals from news streams. Andrée et al (2020) predict the outbreaks of new food crises up to a full year ahead at the administrative level in 21 highrisk countries using the price data from Andrée (2021a,b) combined with conflict data and remote sensing data.…”

“…SEPTEMBER 28, 2022 tail risks up to 3 years ahead. The current paper essentially proposes a middle ground between the work of Wang et al (2020Wang et al ( , 2022 and Andrée et al (2020), combining the multi-year outlook properties provided by the first approach and the predictive optimization of the latter, while aiming for global coverage such as attempted by Martini et al (2021); Foini et al (2022).…”