Data-Driven Ranking of African Food System Transformation Using Dimensionality Reduction

Abstract

Assessing the transformation of food systems across African countries is important for advancing sustainability and economic growth. However, measuring progress remains challenging because of the complexity of indicators and lack of standardized, objective metrics. Existing composite indices often rely on arbitrary weightings that can introduce bias and obscure genuine performance patterns. To address this problem, we present a data-driven ranking approach based on dimensionality reduction and clustering techniques. We construct a country–indicator matrix from binary values, representing whether countries meet specific food system transformation CAADP benchmarks. Next, Principal Component Analysis (PCA) was applied to reduce the dataset’s dimensionality, identifying principal components that reflect countries’ broad progress across domains like agricultural innovation, market integration, nutrition outcomes, environmental sustainability, and governance capacity. Multiple Correspondence Analysis (MCA), which is well-suited to binary data, is also used to help reveal patterns of similarity between countries based on their shared transformation characteristics, offering additional insight that may not be fully captured by PCA. Furthermore, unsupervised learning techniques such as k-means and hierarchical clustering are applied to identify performance groups with similar transformation profiles. This facilitates the classification of countries into groups such as leaders, emerging performers, and laggards, based on underlying structural similarities rather than pre-defined thresholds. Overall, the approach offers a replicable and objective framework for monitoring food system transformation in alignment with the goals of CAADP and Agenda 2063.