A Production-Oriented Machine Learning Pipeline for Real Estate Mass Appraisal with Official Geospatial Feature Engineering

• Gustavo L. B. Pereira UFG / CEIA
• Luís H. A. Rosa UFG / CEIA
• Vinicius Braga UFG / CEIA
• Vinicius Castro UFG / CEIA
• Sofia Costa UFG / CEIA
• Breno M. Barros UFG / CEIA
• Sávio S. T. de Oliveira UFG / CEIA
• Luiz M. L. Pascoal UFG / CEIA

DOI: https://doi.org/10.5753/eniac.2026.31045

Resumo

Real estate mass appraisal underpins fair property taxation in Brazilian municipalities, but attributes that curated-dataset models rely on are unavailable in official cadastral systems at inference time. We present a production-oriented pipeline that restricts features to attributes derivable from the municipal IPTU cadastre, standardizes territory through a spatial join with official shapefiles, and uses a large language model offline to encode neighborhoods into ten locational tiers, persisted as a versioned artifact never invoked at serving time. Among three tree-based ensembles evaluated on about 35,000 deduplicated listings under stratified 5-fold cross-validation with a 90/10 split, CatBoost reaches a holdout mean absolute error near BRL 222,000 to 225,000, R² around 0.89, and a median absolute percentage error close to 11%; coordinates with the language-model group match the full baseline using fewer features.

Referências

Afonso, B. K. d. A., Melo, L. C., Oliveira, W. D. G. d., Sousa, S. B. d. S., and Berton, L. (2019). Housing prices prediction with a deep learning and random forest ensemble. In Anais do XVI Encontro Nacional de Inteligência Artificial e Computacional (ENIAC), pages 389–400. SBC.

Baur, K., Rosenfelder, M., and Lutz, B. (2023). Automated real estate valuation with machine learning models using property descriptions. Expert Systems with Applications, 213:119147.

Bittencourt, L. F., Parraga, O., Ruiz, D. D., Manssour, I. H., Musse, S. R., and Barros, R. C. (2022). Leveraging textual descriptions for house price valuation. In Intelligent Systems: BRACIS 2022, volume 13653 of Lecture Notes in Computer Science, pages 355–369. Springer.

Breiman, L. (2001). Random forests. Machine Learning, 45(1):5–32.

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. ACM.

Faxvaag, H., Aarsnes, U. J. F., and Oust, A. (2026). Incorporating large language models in automated real estate valuation. Journal of Real Estate Research, pages 1–26.

Geerts, M., Reusens, M., Baesens, B., vanden Broucke, S., and De Weerdt, J. (2025). On the performance of LLMs for real estate appraisal. In Machine Learning and Knowledge Discovery in Databases: Applied Data Science Track, pages 195–211. Springer.

Iban, M. C. (2022). An explainable model for the mass appraisal of residences: The application of tree-based ML algorithms and interpretation of value determinants. Habitat International, 128:102660.

Marzagão, T., Ferreira, R., and Sales, L. (2021). A note on real estate appraisal in Brazil. Revista Brasileira de Economia, 75(1):29–36.

Oliveira, A. A. F. d., Bandeira, S. R. V., and Silva, C. V. A. (2022). Estimativa de desempenho de métodos de aprendizado de máquina baseados em árvores de decisão na valoração do solo no município de Fortaleza, Brasil. Revista Valorem, 1(1):1–15.

Prokhorenkova, L., Gusev, G., Vorobev, A., Dorogush, A. V., and Gulin, A. (2018). CatBoost: Unbiased boosting with categorical features. In Advances in Neural Information Processing Systems 31 (NeurIPS), pages 6638–6648.

Rosen, S. (1974). Hedonic prices and implicit markets: Product differentiation in pure competition. Journal of Political Economy, 82(1):34–55.

Sculley, D., Holt, G., Golovin, D., Davydov, E., Phillips, T., Ebner, D., Chaudhary, V., Young, M., Crespo, J.-F., and Dennison, D. (2015). Hidden technical debt in machine learning systems. In Advances in Neural Information Processing Systems 28 (NeurIPS), pages 2503–2511. Curran Associates, Inc.

Sousa Neto, J. N. A. d. and Ladeira, M. (2022). Value estimation of properties administered by the Brazilian army using machine learning and spatial components. In Anais do X Symposium on Knowledge Discovery, Mining and Learning (KDMiLe), pages 17–24. SBC.

Tchuente, D. and Nyawa, S. (2022). Real estate price estimation in French cities using geocoding and machine learning. Annals of Operations Research, 308(1):571–608.

Uriona-Maldonado, M., Vaz, C. R., and Zaghi, L. M. (2024). Real state price estimation in Brazil using machine learning. In Knowledge Management and Artificial Intelligence for Growth, volume 8 of Information Systems Engineering and Management, pages 137–163. Springer.