Machine learning algorithms applied to the estimation of discounts in basic food basket items
DOI:
https://doi.org/10.15675/gepros.3037Keywords:
Machine Learning, Public Administration, Public Procurement, Basic Food Basket, Public PurchasingAbstract
Purpose: This study evaluates machine learning (ML) models for estimating discounts in public procurement of national basic food basket items and analyzes the most relevant features influencing these estimations. Theoretical framework: ML models can be applied in various fields, including engineering, medicine, public health, and economics. These algorithms are capable of uncovering hidden patterns that traditional statistical techniques might miss, making them suitable for predictive and interpretative analyses regarding the characteristics of discounts offered in the procurement of basic food basket items. Methodology/Approach: The algorithms Random Forest, XGBoost, and Artificial Neural Networks were employed and evaluated using mean absolute error and root mean square error. The data were obtained from the Court of Accounts of the State of Paraná and included 18 items from the basic food basket. Findings: Overall, the XGBoost model exhibited the best performance based on error metrics. Regarding feature importance, the ‘quantity’ variable was most significant for estimating discounts on bread and butter, whereas the ‘year of approval’ was a key factor for soybean oil, French bread, beef, rice, and beans. Research, practical & social implications: These models can provide valuable information to managers and oversight bodies, supporting budget planning, fraud detection, and price negotiation in public procurement, thereby contributing to more cost-effective and transparent public administration. However, the models showed limitations in estimating discount values in situations of large variation peaks. Originality/Value: From the preprocessing stage, we suggest a deeper analysis of the data management system, as the fragmented and non-standardized files make data extraction difficult for individuals outside of academia or the corporate sector. ML models can assist public managers in aligning with the principles and trends introduced by the New Public Procurement Law, in a market that accounts for 12% of the national GDP.
References
Aggarwal, C. C. (2015). Data Mining: The textbook. Springer International Publishing. https://doi.org/10.1007/978-3-319-14142-8
Aggarwal, C. C. (2018). Neural Networks and Deep Learning. In Neural Networks and Deep Learning. Springer International Publishing. https://doi.org/10.1007/978-3-319-94463-0
Amorim, V. A. J. de. (2017). Licitações e contratos administrativos: teoria e jurisprudência. https://www2.senado.leg.br/bdsf/handle/id/533714
Bonaccorso, G. (2017). Machine Learning Algorithms: A reference guide to popular algorithms for data science and machine learning (1st ed., Vol. 1). Packt Publishing.
Brasil. (2021). Lei 14133 de Licitações e Contratos Administrativos. https://www.planalto.gov.br/ccivil_03/_ato2019-2022/2021/lei/l14133.htm
Breiman, L. (2001). Random forests. Machine Learning, 45, 5–32. https://doi.org/10.1007/9781441993267_5
Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. https://doi.org/10.1145/2939672.2939785
Dhaliwal, S. S., Nahid, A.-A., & Abbas, R. (2018). Effective Intrusion Detection System Using XGBoost. Information, 9(7), 149. https://doi.org/10.3390/info9070149
França, R. M. M. (2021). Compras públicas conjuntas e economias de escala: um estudo de caso do Sistema COFEN / Conselhos Regionais de Enfermagem [Dissertação de Mestrado, Universidade de Brasília]. http://www.rlbea.unb.br/jspui/handle/10482/42165
Freitas, A. M. de, Prado, F. O., Alexandre, P. L. T., & Carmona, M. F. F. (2021). Nova lei de licitações e contratos administrativos: comentários a lei no 14.133/2021 (2nd ed.). https://www2.senado.leg.br/bdsf/handle/id/588204
Géron, A. (2019). Hands-on: Machine Learning with Scikit-Learn, Keras & Tensorflow (2nd ed.). O’Reilly Media.
Ghazal, M. M., & Hammad, A. (2022). Application of knowledge discovery in database (KDD) techniques in cost overrun of construction projects. International Journal of Construction Management, 22(9), 1632–1646. https://doi.org/10.1080/15623599.2020.1738205
Halužan Vasle, A., & Moškon, M. (2024). Synthetic biological neural networks: From current implementations to future perspectives. BioSystems, 237, 105164. https://doi.org/10.1016/j.biosystems.2024.105164
Instituto Brasileiro de Geografia e Estatística. (2024). Sistema Nacional de Índices de Preços ao Consumidor. Tabela 7063 - INPC. https://sidra.ibge.gov.br/tabela/7063
Lara, J. A., Lizcano, D., Martínez, M. A., & Pazos, J. (2014). Data preparation for KDD through automatic reasoning based on description logic. Information Systems, 44, 54–72. https://doi.org/10.1016/j.is.2014.03.002
M. Mijwil, M. (2021). Artificial Neural Networks Advantages and Disadvantages. Mesopotamian Journal of Big Data, 2021, 29–31. https://doi.org/10.58496/MJBD/2021/006
Mello, C. A. B. de. (2015). Curso de Direito Administrativo (32nd ed.). Malheiros.
Morabito, R., & Pureza, V. (2018). Modelagem e Simulação. In Cauchick, P.A., (Coord.). Metodologia de Pesquisa em Engenharia de Produção e Gestão de Operações (3 ed., Chap. 8, pp. 165-195). Rio de Janeiro: Elsevier.
Oliveira, L. H. R. de, Rêgo, T. G. do, & Diniz, J. A. (2019). Previsão de Valores de Aquisições Governamentais: o Uso dos Conceitos de Data Science e Machine Learning. XVI Congresso USP de Iniciação Científica Em Contabilidade, 1–15.
Radhoush, S., Whitaker, B. M., & Nehrir, H. (2023). An Overview of Supervised Machine Learning Approaches for Applications in Active Distribution Networks. Energies, 16(16), 5972. https://doi.org/10.3390/en16165972
Raschka, S. (2015). Python Machine Learning (1st ed.). Packt Publishing Ltd.
Ribeiro, C. G., & Inácio Júnior, E. (2019). O mercado de compras governamentais brasileiro (2006-2017): Mensuração e análise.
Sagi, O., & Rokach, L. (2021). Approximating XGBoost with an interpretable decision tree. Information Sciences, 572, 522–542. https://doi.org/10.1016/j.ins.2021.05.055
Sampaio, I. G., Bernardini, F., Paes, A., Andrade, E. de O., & Viterbo, J. (2019). Avaliação de Modelos de Predição e Previsão Construídos por Algoritmos de Aprendizado de Máquina em Problemas de Cidades Inteligentes. In Tópicos em Sistemas de Informação: Minicursos SBSI 2019 (pp. 81–113). SBC. https://doi.org/10.5753/sbc.480.9.04
Santos, L. B., Gentry, D., Tryforos, A., Fultz, L., Beasley, J., & Gentimis, T. (2024). Soybean yield prediction using machine learning algorithms under a cover crop management system. Smart Agricultural Technology, 8, 100442. https://doi.org/10.1016/j.atech.2024.100442
Signor, R., Marchiori, F. F., Raupp, A. B., Magro, R. R., & Lopes, A. de O. (2022). A nova lei de licitações como promotora da maldição do vencedor. Revista de Administração Pública, 56(1), 176–190. https://doi.org/10.1590/0034-761220210133
Silva, M. O., Costa, L. L., Bezerra, G., Gomide, L. D., Hott, H. R., Oliveira, G. P., Brandão, M. A., Lacerda, A., & Pappa, G. (2023). Análise de Sobrepreço em Itens de Licitações Públicas. Anais Do XI Workshop de Computação Aplicada Em Governo Eletrônico (WCGE 2023), 118–129. https://doi.org/10.5753/wcge.2023.230608
Tan, P.-N., Steinbach, M., Karpatne, A., & Kumar, V. (2019). Introduction to Data Mining (2nd ed.). Pearson Prentice Hall.
Willmott, C. J., & Matsuura, K. (2005). Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Climate Research, 30(1), 79–82. http://www.jstor.org/stable/24869236
Xerez, S. R. D. (2013). A evolução do orçamento público e seus instrumentos de planejamento. Revista Científica Semana Acadêmica, 01(43), 1–19.
Xu, Q., & Yin, J. (2021). Application of Random Forest Algorithm in Physical Education. Scientific Programming, 2021. https://doi.org/10.1155/2021/1996904
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2025 Jose Mateus Rodrigues dos Santos, Bruno Samways dos Santos

This work is licensed under a Creative Commons Attribution 4.0 International License.
O(s) autor(es) do artigo autorizam a publicação do texto na revista e garantem que a contribuição é original e inédita, não estando em processo de avaliação em outra(s) revista(s). As opiniões, ideias e conceitos emitidos nos textos são de inteira responsabilidade do(s) autor(es), não sendo a revista responsável por tais conteúdos.
Os editores da revista reservam o direito de efetuar ajustes textuais e de adequação às normas da publicação, caso necessário.
Os autores mantêm os direitos autorais sobre o trabalho e concedem à revista o direito de primeira publicação, sendo o trabalho simultaneamente licenciado sob a Attribution 4.0 International (CC BY 4.0), o que permite o compartilhamento do trabalho com reconhecimento da autoria e publicação inicial nesta revista.
Os autores têm autorização para firmar contratos adicionais, separadamente, para distribuição não-exclusiva da versão do trabalho publicada nesta revista (ex.: publicar em repositório institucional ou como capítulo de livro), com reconhecimento de autoria e publicação inicial nesta revista.


