GerModel: Gestão de Modelos em Dados em Fluxo com Conceitos Recorrentes
Resumo
Fluxos meteorológicos de alta frequência apresentam não estacionariedade e concept drift recorrente, o que dificulta a manutenção de modelos preditivos. Este trabalho formula esse cenário como um problema de gestão de modelos em fluxo, no qual se decide entre retreinamento completo e reutilização de modelos históricos. Para isso, apresenta-se a GerModel, uma abordagem preliminar baseada no Population Stability Index (PSI) para selecionar e reutilizar modelos históricos em previsão horária de precipitação. A avaliação usa dados do aeroporto de Miami entre 2001 e 2025, com processamento mensal, e compara a GerModel com o retreinamento completo após mudança. No cenário avaliado, os resultados sustentam que a abordagem pode constituir alternativa viável, com F1 competitivo, alto reuso e menor custo computacional.
Palavras-chave:
Dados em Fluxo, Deriva de Conceito, Reutilização de Modelos, Gestão de Modelos, Aprendizado Adaptativo
Referências
Anderson, R., Koh, Y. S., Dobbie, G., and Bifet, A. (2019). Recurring concept meta-learning for evolving data streams. Expert Systems with Applications, 138:112832.
Bayram, F., Ahmed, B. S., and Kassler, A. (2022). From concept drift to model degradation: An overview on performance-aware drift detectors. Knowledge-Based Systems, 245.
Gama, J., Zliobaite, I., Bifet, A., Pechenizkiy, M., and Bouchachia, A. (2014). A survey on concept drift adaptation. ACM Computing Surveys, 46.
Halstead, B., Koh, Y. S., Riddle, P., Pears, R., Pechenizkiy, M., and Bifet, A. (2021). Recurring concept memory management in data streams: exploiting data stream concept evolution to improve performance and transparency. Data Mining and Knowledge Discovery, 35:796 – 836.
Iowa State University Mesonet (2026). FL ASOS Data Download. Technical report.
Kurian, J. F. and Allali, M. (2024). Detecting drifts in data streams using Kullback-Leibler (KL) divergence measure for data engineering applications. Journal of Data, Information and Management, 6:207 – 216.
Lu, S.-C., Song, W., Pfob, A., and Gibbons, C. (2025). Assessing the representativeness of large medical data using population stability index. BMC Medical Research Methodology, 25.
Misra, V. and Jayasankar, C. (2024). Advancing the Seasonal Outlook of the Wet Seasons of Florida. Weather and Forecasting, 39:1751 – 1759.
Ogasawara, E., Castro, A., Salles, D., Lima, J., Tavares, L., Carvalho, D., Bezerra, E., and Coutinho, R. (2025a). daltoolbox: Leveraging Experiment Lines to Data Analytics.
Ogasawara, E., Salles, R., Porto, F., and Pacitti, E. (2025b). Event Detection in Time Series. Synthesis Lectures on Data Management. Springer Nature Switzerland, Cham, 1 edition.
Ortíz Díaz, A., Del Campo-Ávila, J., Ramos-Jiménez, G., Frías Blanco, I., Caballero Mota, Y., Mustelier Hechavarría, A., and Morales-Bueno, R. (2015). Fast adapting ensemble: A new algorithm for mining data streams with concept drift. Scientific World Journal, 2015:235810.
Peeling, J. A., Judge, J., Misra, V., Jayasankar, C., and Lusher, W. R. (2023). Gap-free 16-year (2005–2020) sub-diurnal surface meteorological observations across Florida. Scientific Data, 10.
Peres, F. F. (2026). Effect sizes for nonparametric tests. Biochemia Medica, 36.
Read, J. and Zliobaite, I. (2025). Supervised Learning from Data Streams: An Overview and Update. ACM Computing Surveys, 57:1–37.
Sun, Y., Wang, Z., Bai, Y., Dai, H., and Nahavandi, S. (2018). A Classifier Graph Based Recurring Concept Detection and Prediction Approach. Computational Intelligence and Neuroscience, 2018:4276291.
Tavares, L., Carvalho, L., Machado, R., Carvalho, D., Pacitti, E., Porto, F., and Ogasawara, E. (2025). heimdall: Drift Adaptable Models.
Webb, G. I., Hyde, R., Cao, H., Nguyen, H. L., and Petitjean, F. (2016). Characterizing concept drift. Data Mining and Knowledge Discovery, 30:964 – 994.
Yurdakul, B. and Naranjo, J. (2020). Statistical properties of the population stability index. Journal of Risk Model Validation, 14:89 – 100.
Zhao, P., Cai, L.-W., and Zhou, Z.-H. (2020). Handling concept drift via model reuse. Machine Learning, 109:533 – 568.
Bayram, F., Ahmed, B. S., and Kassler, A. (2022). From concept drift to model degradation: An overview on performance-aware drift detectors. Knowledge-Based Systems, 245.
Gama, J., Zliobaite, I., Bifet, A., Pechenizkiy, M., and Bouchachia, A. (2014). A survey on concept drift adaptation. ACM Computing Surveys, 46.
Halstead, B., Koh, Y. S., Riddle, P., Pears, R., Pechenizkiy, M., and Bifet, A. (2021). Recurring concept memory management in data streams: exploiting data stream concept evolution to improve performance and transparency. Data Mining and Knowledge Discovery, 35:796 – 836.
Iowa State University Mesonet (2026). FL ASOS Data Download. Technical report.
Kurian, J. F. and Allali, M. (2024). Detecting drifts in data streams using Kullback-Leibler (KL) divergence measure for data engineering applications. Journal of Data, Information and Management, 6:207 – 216.
Lu, S.-C., Song, W., Pfob, A., and Gibbons, C. (2025). Assessing the representativeness of large medical data using population stability index. BMC Medical Research Methodology, 25.
Misra, V. and Jayasankar, C. (2024). Advancing the Seasonal Outlook of the Wet Seasons of Florida. Weather and Forecasting, 39:1751 – 1759.
Ogasawara, E., Castro, A., Salles, D., Lima, J., Tavares, L., Carvalho, D., Bezerra, E., and Coutinho, R. (2025a). daltoolbox: Leveraging Experiment Lines to Data Analytics.
Ogasawara, E., Salles, R., Porto, F., and Pacitti, E. (2025b). Event Detection in Time Series. Synthesis Lectures on Data Management. Springer Nature Switzerland, Cham, 1 edition.
Ortíz Díaz, A., Del Campo-Ávila, J., Ramos-Jiménez, G., Frías Blanco, I., Caballero Mota, Y., Mustelier Hechavarría, A., and Morales-Bueno, R. (2015). Fast adapting ensemble: A new algorithm for mining data streams with concept drift. Scientific World Journal, 2015:235810.
Peeling, J. A., Judge, J., Misra, V., Jayasankar, C., and Lusher, W. R. (2023). Gap-free 16-year (2005–2020) sub-diurnal surface meteorological observations across Florida. Scientific Data, 10.
Peres, F. F. (2026). Effect sizes for nonparametric tests. Biochemia Medica, 36.
Read, J. and Zliobaite, I. (2025). Supervised Learning from Data Streams: An Overview and Update. ACM Computing Surveys, 57:1–37.
Sun, Y., Wang, Z., Bai, Y., Dai, H., and Nahavandi, S. (2018). A Classifier Graph Based Recurring Concept Detection and Prediction Approach. Computational Intelligence and Neuroscience, 2018:4276291.
Tavares, L., Carvalho, L., Machado, R., Carvalho, D., Pacitti, E., Porto, F., and Ogasawara, E. (2025). heimdall: Drift Adaptable Models.
Webb, G. I., Hyde, R., Cao, H., Nguyen, H. L., and Petitjean, F. (2016). Characterizing concept drift. Data Mining and Knowledge Discovery, 30:964 – 994.
Yurdakul, B. and Naranjo, J. (2020). Statistical properties of the population stability index. Journal of Risk Model Validation, 14:89 – 100.
Zhao, P., Cai, L.-W., and Zhou, Z.-H. (2020). Handling concept drift via model reuse. Machine Learning, 109:533 – 568.
Publicado
08/09/2026
Como Citar
MACHADO, Rodrigo; TAVARES, Lucas Giusti; BARONI, Lais; PORTO, Fabio; COUTINHO, Rafaelli de Carvalho; OGASAWARA, Eduardo.
GerModel: Gestão de Modelos em Dados em Fluxo com Conceitos Recorrentes. In: SIMPÓSIO BRASILEIRO DE BANCO DE DADOS (SBBD), 41. , 2026, São Carlos/SP.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2026
.
p. 826-832.
ISSN 2763-8979.
DOI: https://doi.org/10.5753/sbbd.2026.249419.
