A Machine Learning Framework for Early Detection of Food Insecurity Using Administrative Microdata

• Keila Barbosa UFAL / UFPE
• Andre L. Aquino UFAL

DOI: https://doi.org/10.5753/lasdigov.2026.23871

Resumo

Food insecurity is a multidimensional phenomenon strongly associated with socioeconomic, household, and territorial factors. This study proposes a machine learning pipeline to predict food insecurity risk using administrative microdata from Brazil’s Cadastro Único (CadÚnico) from the state of Alagoas. The approach integrates feature engineering, predictive modeling, temporal evaluation, SHAP-based interpretability, and spatial aggregation of predictions. We evaluate both linear and tree-based models, including Logistic Regression, Random Forest, LightGBM, and CatBoost. A temporal evaluation protocol is adopted, using 2024 data for training/validation and 2025 data for testing, improving temporal generalization and evaluation realism. The final LightGBM model achieved an ROC-AUC of 0.91 and PR-AUC of 0.72 on the temporal test set, with a recall of 0.78, indicating strong capability in identifying at-risk households. Interpretability analysis highlights the importance of territorial and socioeconomic variables, while spatial aggregation reveals clear geographic patterns of vulnerability. The results demonstrate the potential of combining administrative data and machine learning to support early warning systems and inform data-driven public policies.

Referências

Alkire, S., Roche, J. M., Ballon, P., Foster, J., Santos, M. E., and Seth, S. (2015). Multidimensional poverty measurement and analysis. Oxford University Press, USA.

Anselin, L. (1995). Local indicators of spatial association—lisa. Geographical Analysis, 27(2):93–115.

Chen, T. and Guestrin, C. (2016). Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, pages 785–794.

Einav, L. and Levin, J. (2014). Economics in the age of big data. Science, 346(6210):1243089.

Grinsztajn, L., Oyallon, E., and Varoquaux, G. (2022). Why do tree-based models still outperform deep learning on typical tabular data? Advances in neural information processing systems, 35:507–520.

Haixiang, G., Yijing, L., Shang, J., Mingyun, G., Yuanyue, H., and Bing, G. (2017). Learning from class-imbalanced data: Review of methods and applications. Expert systems with applications, 73:220–239.

He, H. and Garcia, E. A. (2009). Learning from imbalanced data. IEEE Transactions on Knowledge and Data Engineering, 21(9):1263–1284.

Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., and Liu, T.-Y. (2017). Lightgbm: A highly efficient gradient boosting decision tree. Advances in neural information processing systems, 30.

Lundberg, S. M. and Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in neural information processing systems, 30.

Saito, T. and Rehmsmeier, M. (2015). The precision-recall plot is more informative than the roc plot when evaluating binary classifiers on imbalanced datasets. PLOS ONE, 10(3):e0118432.

Sen, A. (1982). Poverty and famines: An essay on entitlement and deprivation. Oxford university press.

UNICEF et al. (2021). The state of food security and nutrition in the world 2021. FAO;.