Highlights:

• A Food Security Identification (FSI) roadmap is proposed for the development of a multi-dimensional food security index.

• The roadmap employs a machine learning approach towards identifying regions which are estimated to be food insecure.

• The practicality of the proposed roadmap is demonstrated via a case study pertaining to South Africa.

Abstract:

Food insecurity is a multifaceted problem. It is, in fact, one of the most significant concerns of the 21st century. According to the United Nations, nearly one in every three people, or 2.3 billion people worldwide, experienced a moderate to severe degree of food insecurity in 2021. As a result, there is a rising recognition of the critical need for successfully identifying, monitoring, and improving the food security of vulnerable populations. An effective method for anticipating the degree to which a region may be considered food insecure, however, is required before attempting to launch remedies for such a situation. To this end, we propose a novel machine learning-based spatial index in this article for estimating the degree of food insecurity experienced by a population. A real-world case study pertaining to South Africa is presented which demonstrates the applicability of our modelling approach. In particular, our roadmap designed towards developing the index is based on a spatial analysis of the study region and the construction of an appropriate data set. Following that, the strength of machine learning algorithms is harnessed to estimate the risk of food insecurity. Finally, the results of the aforementioned components are analysed to determine the causes and features of food insecure regions. The demonstration of the roadmap presented in this article exhibits promising results in its ability to classify 92.3% of food-insecure regions in South Africa correctly, thereby potentially supporting decision makers in their quest to improve food security policies and programmes. The research presented here may, in future, be extended by including additional descriptive features in the data set, increasing decision support utility through actionable recommendations, and exploring the temporal aspect of food security by incorporating cross-sectional data.