This study used data from 13,136 Nigerian children aged 6 to 59 months to examine what drives the country’s high rates of childhood anaemia. The researchers tested 16 different machine learning algorithms and found that the Extra Trees classifier produced the strongest results, with an AUC of 0.8319 and an accuracy and recall of 0.7565. After selecting this model, they ranked the most influential predictors, which included household structure such as the number of under five children, birth order, and the child’s age. Maternal behaviors, especially health-seeking patterns, also played a major role which suggests that caregiver actions can alter risks. The model flagged several socioeconomic pressures like money problems and limited media access as additional contributors. Environmental features, including proximity to water, land-surface temperature, and local population density, appeared to be surprisingly impactful. To test fairness, the authors looked at how the model performed across regions, wealth groups, ethnicities, and gender. They found that accuracy varied, with the lowest AUCs reported in the northeast 0.79, the poorest wealth index category 0.80, and the Hausa/Fulani group 0.81. Overall, the study argues that machine learning can help highlight high risk clusters and guide more strategic public health efforts.