Food Basket Optimization for Nigerian Households

Nigeria is facing a severe food insecurity situation, with millions of people across the country struggling to access sufficient and nutritious food. The causes of food insecurity in Nigeria are complex and multi-faceted. Although the Nigerian government and international aid agencies have taken steps to address food insecurity, the scale of the problem remains significant and more needs to be done to alleviate food insecurity in Nigeria. To address the food insecurity issue for Nigerian households, we propose an additional tool for the government’s toolbox: an optimal food basket designed to be affordable and satisfy food consumption patterns of Nigerian households. Our project proposes the use of a random forest supervised learning model to predict the food insecurity levels of households (low, medium or high) based on a food insecurity indicator (Coping Strategies Index). Alternatively, if households' food insecurity status is known, we propose the use of a K-Means clustering unsupervised learning model to group Nigerian households into clusters based on their characteristics (including food coping behaviors). Our specific model suggests two clusters. The machine learning models are used to obtain median food expenditure for each household group, which is then used to design two food basket models: a low-cost and a low-cost & culturally-acceptable food basket. We found that the low-cost food basket model cost NGN 115 across household groups and comprised 4 food items. While the low-cost & culturally-acceptable food basket model comprised a wider variety of food items and was priced at the median food expenditure for each household group. As a means to fight food insecurity, we recommend the food basket model optimized for low-cost and culturally acceptability to the Nigerian government for implementation in hunger alleviation, cash transfer and health education programs. We acknowledge that our project is subject to several data, analytical and economic limitations, some of which we address as future work.