Evaluation of Machine Learning Algorithms to Estimate the Prevalence of Anemia in Children at the District Level in Peru

Mamani A.; Gutierrez-Pachas D.A.; Rosero L. · Communications in Computer and Information Science (2026) · DOI: 10.1007/978-3-032-20322-9_33

Resumen

An adequate assessment of public health indicators and metrics is crucial for national governments, enabling them to formulate more efficient public policies that benefit society. Traditionally, statistical methods were employed to obtain spatial estimates in small areas with limited or no information, a technique known as Small Area Estimation (SAE). Despite the success of these statistical techniques, they present challenges when dealing with increasing dimensionality and volume of information. For this reason, evaluating modern techniques is crucial to assessing SAE problems. We address an alternative to traditional methods for SAE problems using Machine Learning (ML) algorithms. To validate our results, we utilize the estimation of childhood anemia prevalence at the district level in Peru, based on real data, and compare it with the statistical method, spatial Fay-Herriot. Our evaluation uses regression and autocorrelation spatial metrics, as well as consistency with real data. Our results demonstrate that ML algorithms can offer competitive, and in some cases, superior, predictive capabilities despite data scarcity and inconsistencies. Finally, we conclude that ML algorithms are a viable alternative for governments to establish more efficient public policies. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.

Instituciones: Universidad Católica San Pablo