A methodology for synthetic data generation and validation applied to agricultural engineering

Detalhes bibliográficos
Ano de defesa: 2026
Autor(a) principal: Brandão Junior, Luiz Carlos lattes
Outros Autores: https://orcid.org/0009-0004-8513-0863
Orientador(a): Magalhães, Ricardo Rodrigues
Banca de defesa: Não Informado pela instituição
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: eng
Instituição de defesa: Universidade Federal de Lavras
Escola de Engenharia (EENG)
Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia Agrícola
Departamento: Não Informado pela instituição
País: brasil
Palavras-chave em Português:
Área do conhecimento CNPq:
Link de acesso: https://repositorio.ufla.br/handle/1/61086
Resumo: The scarcity of high-quality experimental data, coupled with the high costs of laboratory and sensory analyses, are significant barriers to the development of computational modeling and artificial intelligence in the field of Agricultural Sciences. This thesis introduces a reverse engineering framework for generating synthetic data based on Cholesky decomposition, which can reconstruct multivariate distributions while preserving statistical properties and structural dependence between variables. As a case study, the technique was employed in a model for the sensory quality of Arabica coffee using a public dataset from the Coffee Quality Institute ( = 207 samples). The method was divided into four stages: (i) data extraction and preprocessing (ii) statistical synthesis preserving first and second order moments (iii) multiple validation of statistical fidelity and (iv) evaluation of predictive power via the Train Synthetic, Test Real (TSTR) protocol. The results showed high structural fidelity, obtaining a correlation matrix similarity of 96.20% and a Kolmogorov-Smirnov (KS) test similarity of 85.81%; all variables exceeded the technical acceptance threshold of 70%. In RF regression, only the models trained with synthetic data achieved R² = 0.9498, a Mean Absolute Error (MAE) of only 0.216, and 100% accuracy within a margin of error of ±10%. For the ordinal classification of sensory quality, the approach achieved 89.37% accuracy and a Weighted Quadratic Kappa Coefficient of 0.9558, showing almost perfect agreement with the standards of the Specialty Coffee Association (SCA). As its main scientific contribution, this thesis establishes a reverse engineering framework for reconstructing synthetic databases from aggregated statistics, demonstrating that the simultaneous preservation of statistical fidelity and predictive utility can be achieved without the use of deep generative models. The proposed approach expands the possibilities for secure data sharing, reduces dependence on costly experimentation, and offers a computationally efficient and easily reproducible solution with high potential for technology transfer to applications in Agricultural Engineering, Agronomy, and other areas that depend on reliable experimental databases.
id UFLA_350a41efb85d9e242969cb9bd3bf5cd2
oai_identifier_str oai:repositorio.ufla.br:1/61086
network_acronym_str UFLA
network_name_str Repositório Institucional da UFLA
repository_id_str
spelling Ferreira, Danton DiegoFerreira, Danton DiegoLima, Danilo Alves deLacerda, Wilian SoaresVitor, Giovanni BernardesMagalhães, Ricardo Rodrigueshttp://lattes.cnpq.br/4992193748825216Brandão Junior, Luiz Carloshttps://orcid.org/0009-0004-8513-08632026-08-14T20:37:42Z2026-07-09BRANDÃO JÚNIOR, Luiz Carlos. A methodology for synthetic data generation and validation applied to agricultural engineering. 2026. 172 p. Tese (Doutorado em Engenharia Agrícola) - Universidade Federal de Lavras, Lavras, 2026.https://repositorio.ufla.br/handle/1/61086The scarcity of high-quality experimental data, coupled with the high costs of laboratory and sensory analyses, are significant barriers to the development of computational modeling and artificial intelligence in the field of Agricultural Sciences. This thesis introduces a reverse engineering framework for generating synthetic data based on Cholesky decomposition, which can reconstruct multivariate distributions while preserving statistical properties and structural dependence between variables. As a case study, the technique was employed in a model for the sensory quality of Arabica coffee using a public dataset from the Coffee Quality Institute ( = 207 samples). The method was divided into four stages: (i) data extraction and preprocessing (ii) statistical synthesis preserving first and second order moments (iii) multiple validation of statistical fidelity and (iv) evaluation of predictive power via the Train Synthetic, Test Real (TSTR) protocol. The results showed high structural fidelity, obtaining a correlation matrix similarity of 96.20% and a Kolmogorov-Smirnov (KS) test similarity of 85.81%; all variables exceeded the technical acceptance threshold of 70%. In RF regression, only the models trained with synthetic data achieved R² = 0.9498, a Mean Absolute Error (MAE) of only 0.216, and 100% accuracy within a margin of error of ±10%. For the ordinal classification of sensory quality, the approach achieved 89.37% accuracy and a Weighted Quadratic Kappa Coefficient of 0.9558, showing almost perfect agreement with the standards of the Specialty Coffee Association (SCA). As its main scientific contribution, this thesis establishes a reverse engineering framework for reconstructing synthetic databases from aggregated statistics, demonstrating that the simultaneous preservation of statistical fidelity and predictive utility can be achieved without the use of deep generative models. The proposed approach expands the possibilities for secure data sharing, reduces dependence on costly experimentation, and offers a computationally efficient and easily reproducible solution with high potential for technology transfer to applications in Agricultural Engineering, Agronomy, and other areas that depend on reliable experimental databases.A escassez de dados experimentais de qualidade, juntamente com os altos custos de análises laboratoriais e sensoriais, são barreiras significativas para o desenvolvimento da modelagem computacional e da inteligência artificial na área das Ciências Agrárias. Esta tese introduz um framework de engenharia reversa para a geração de dados sintéticos baseado na decomposição de Cholesky, que pode reconstruir distribuições multivariadas enquanto preserva as propriedades estatísticas e a dependência estrutural entre as variáveis. Como caso de estudo, a técnica foi empregada em um modelo para a qualidade sensorial do café Arábica com um conjunto de dados público do Coffee Quality Institute ( = 207 amostras). O método foi dividido em quatro estágios: (i) extração e pré-processamento de dados (ii) síntese estatística preservando momentos de ordem um e dois (iii) validação múltipla de fidelidade estatística e (iv) avaliação do poder preditivo via o protocolo Train Synthetic, Test Real (TSTR). Os resultados mostraram uma elevada fidelidade estrutural, obtendo-se similaridade da matriz de correlação 96,20% e similaridade pelo teste de Kolmogorov-Smirnov (KS) 85.81%, todas as variáveis ultrapassaram o limiar técnico de aceitação de 70%. Na regressão RF, apenas os modelos treinados com dados sintéticos alcançaram 2 = 0, 9498, Erro Médio Absoluto (MAE) de apenas 0,216 e precisão de 100% dentro da margem de erro de ±10%. Para a classificação ordinal da qualidade sensorial, a abordagem atingiu 89,37% de acurácia e (Weighted Quadratic) Kappa Coeficiente de 0,9558, mostrando concordância quase perfeita com os padrões da Specialty Coffee Asscociation (SCA). Como principal contribuição científica, esta tese estabelece um framework de engenharia reversa para reconstrução de bases de dados sintéticas a partir de estatísticas agregadas, demonstrando que a preservação simultânea da fidelidade estatística e da utilidade preditiva pode ser alcançada sem o emprego de modelos generativos profundos. A abordagem proposta amplia as possibilidades de compartilhamento seguro de dados, reduz a dependência de experimentação onerosa e oferece uma solução computacionalmente eficiente e de fácil reprodução, com elevado potencial de transferência tecnológica para aplicações em Engenharia Agrícola, Agronomia e demais áreas que dependem de bases de dados experimentais confiáveis.Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)Máquinas e Mecanização AgrícolaSociaisTecnológicoEconômicosMeio ambienteSaúdeTecnologia e produçãoODS 2: Fome zero e agricultura sustentávelODS 3: Saúde e bem-estarODS 7: Energia limpa e acessívelODS 9: Indústria, inovação e infraestruturaODS 12: Consumo e produção responsáveisODS 13: Ação contra a mudança global do climaODS 15: Vida terrestre2. Uso sustentável da água e do solo6. Eficiência energética ou inovação ambientalEngenharia AgrícolaUniversidade Federal de LavrasEscola de Engenharia (EENG)Programa de Pós-Graduação em Engenharia AgrícolaUFLAbrasilAttribution 3.0 Brazilhttp://creativecommons.org/licenses/by/3.0/br/info:eu-repo/semantics/openAccessEngenharia AgrícolaSynthetic dataCholesky decompositionArabica coffeeSensory analysisMachine learningCafé Arábica - Análise sensorialAnálise sensorialDados sintéticosAprendizado de máquinaA methodology for synthetic data generation and validation applied to agricultural engineeringUma metodologia para geração e validação de dados sintéticos aplicada à engenharia agrícolainfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesisengreponame:Repositório Institucional da UFLAinstname:Universidade Federal de Lavras (UFLA)instacron:UFLACC-LICENSElicense_rdflicense_rdfapplication/rdf+xml; charset=utf-81025https://repositorio.ufla.br/bitstreams/47d6fd07-6439-4e54-97ae-77a644f1b384/download5a033ee506f3a0a175bee8fc81f0bd66MD51falseAnonymousREADLICENSElicense.txtlicense.txttext/plain; charset=utf-8955https://repositorio.ufla.br/bitstreams/af253796-395e-4b46-9117-c1a0717842b7/downloaddc1a173fe9489e283d3a1f54f6ab2ab9MD52falseAnonymousREADORIGINALTexto completoTexto completoapplication/pdf19260992https://repositorio.ufla.br/bitstreams/6d64f8fa-7bc2-4955-9be5-39e327139fda/download23e5b2419024f37919fe36ecce4eebf7MD54trueAnonymousREADImpactos da pesquisaImpactos da pesquisaapplication/pdf604930https://repositorio.ufla.br/bitstreams/403d9f45-e622-4419-878d-edc1a7cf8703/downloadd13b9ee9893f27a614510016898db2a7MD55falseAnonymousREADTEXTTexto completo.txtTexto completo.txtExtracted texttext/plain102700https://repositorio.ufla.br/bitstreams/0e649d04-d9a8-4f26-8ee8-601fbc8e33e0/downloadc85082449256160eff49955d38f1f236MD56falseAnonymousREADImpactos da pesquisa.txtImpactos da pesquisa.txtExtracted texttext/plain7060https://repositorio.ufla.br/bitstreams/af65f522-019d-4383-b711-b2da5feec9d2/downloadfe3206374123db5d1ff9bf27133cc2b0MD58falseAnonymousREADTHUMBNAILTexto completo.jpgTexto completo.jpgGenerated Thumbnailimage/jpeg2817https://repositorio.ufla.br/bitstreams/6325e64f-e4bf-41f1-a2d8-a9823698e6b6/downloadaf20e4165ca68393f1235565f39a7967MD57falseAnonymousREADImpactos da pesquisa.jpgImpactos da pesquisa.jpgGenerated Thumbnailimage/jpeg4375https://repositorio.ufla.br/bitstreams/bdb52171-2f84-4563-a4ed-549dc91604e0/download361f7d26001ffad8bb602dc4a509c70dMD59falseAnonymousREAD1/610862026-09-08 12:43:36.341http://creativecommons.org/licenses/by/3.0/br/Attribution 3.0 Brazilopen.accessoai:repositorio.ufla.br:1/61086https://repositorio.ufla.brRepositório InstitucionalPUBhttps://repositorio.ufla.br/server/oai/requestnivaldo@ufla.br || repositorio.biblioteca@ufla.bropendoar:2026-09-08T15:43:36Repositório Institucional da UFLA - Universidade Federal de Lavras (UFLA)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
dc.title.none.fl_str_mv A methodology for synthetic data generation and validation applied to agricultural engineering
dc.title.alternative.none.fl_str_mv Uma metodologia para geração e validação de dados sintéticos aplicada à engenharia agrícola
title A methodology for synthetic data generation and validation applied to agricultural engineering
spellingShingle A methodology for synthetic data generation and validation applied to agricultural engineering
Brandão Junior, Luiz Carlos
Engenharia Agrícola
Synthetic data
Cholesky decomposition
Arabica coffee
Sensory analysis
Machine learning
Café Arábica - Análise sensorial
Análise sensorial
Dados sintéticos
Aprendizado de máquina
title_short A methodology for synthetic data generation and validation applied to agricultural engineering
title_full A methodology for synthetic data generation and validation applied to agricultural engineering
title_fullStr A methodology for synthetic data generation and validation applied to agricultural engineering
title_full_unstemmed A methodology for synthetic data generation and validation applied to agricultural engineering
title_sort A methodology for synthetic data generation and validation applied to agricultural engineering
author Brandão Junior, Luiz Carlos
author_facet Brandão Junior, Luiz Carlos
https://orcid.org/0009-0004-8513-0863
author_role author
author2 https://orcid.org/0009-0004-8513-0863
author2_role author
dc.contributor.co-advisor.none.fl_str_mv Ferreira, Danton Diego
dc.contributor.referee.none.fl_str_mv Ferreira, Danton Diego
Lima, Danilo Alves de
Lacerda, Wilian Soares
Vitor, Giovanni Bernardes
dc.contributor.advisor1.fl_str_mv Magalhães, Ricardo Rodrigues
dc.contributor.authorLattes.fl_str_mv http://lattes.cnpq.br/4992193748825216
dc.contributor.author.fl_str_mv Brandão Junior, Luiz Carlos
https://orcid.org/0009-0004-8513-0863
contributor_str_mv Magalhães, Ricardo Rodrigues
dc.subject.cnpq.fl_str_mv Engenharia Agrícola
topic Engenharia Agrícola
Synthetic data
Cholesky decomposition
Arabica coffee
Sensory analysis
Machine learning
Café Arábica - Análise sensorial
Análise sensorial
Dados sintéticos
Aprendizado de máquina
dc.subject.por.fl_str_mv Synthetic data
Cholesky decomposition
Arabica coffee
Sensory analysis
Machine learning
Café Arábica - Análise sensorial
Análise sensorial
Dados sintéticos
Aprendizado de máquina
description The scarcity of high-quality experimental data, coupled with the high costs of laboratory and sensory analyses, are significant barriers to the development of computational modeling and artificial intelligence in the field of Agricultural Sciences. This thesis introduces a reverse engineering framework for generating synthetic data based on Cholesky decomposition, which can reconstruct multivariate distributions while preserving statistical properties and structural dependence between variables. As a case study, the technique was employed in a model for the sensory quality of Arabica coffee using a public dataset from the Coffee Quality Institute ( = 207 samples). The method was divided into four stages: (i) data extraction and preprocessing (ii) statistical synthesis preserving first and second order moments (iii) multiple validation of statistical fidelity and (iv) evaluation of predictive power via the Train Synthetic, Test Real (TSTR) protocol. The results showed high structural fidelity, obtaining a correlation matrix similarity of 96.20% and a Kolmogorov-Smirnov (KS) test similarity of 85.81%; all variables exceeded the technical acceptance threshold of 70%. In RF regression, only the models trained with synthetic data achieved R² = 0.9498, a Mean Absolute Error (MAE) of only 0.216, and 100% accuracy within a margin of error of ±10%. For the ordinal classification of sensory quality, the approach achieved 89.37% accuracy and a Weighted Quadratic Kappa Coefficient of 0.9558, showing almost perfect agreement with the standards of the Specialty Coffee Association (SCA). As its main scientific contribution, this thesis establishes a reverse engineering framework for reconstructing synthetic databases from aggregated statistics, demonstrating that the simultaneous preservation of statistical fidelity and predictive utility can be achieved without the use of deep generative models. The proposed approach expands the possibilities for secure data sharing, reduces dependence on costly experimentation, and offers a computationally efficient and easily reproducible solution with high potential for technology transfer to applications in Agricultural Engineering, Agronomy, and other areas that depend on reliable experimental databases.
publishDate 2026
dc.date.accessioned.fl_str_mv 2026-08-14T20:37:42Z
dc.date.issued.fl_str_mv 2026-07-09
dc.type.status.fl_str_mv info:eu-repo/semantics/publishedVersion
dc.type.driver.fl_str_mv info:eu-repo/semantics/masterThesis
format masterThesis
status_str publishedVersion
dc.identifier.citation.fl_str_mv BRANDÃO JÚNIOR, Luiz Carlos. A methodology for synthetic data generation and validation applied to agricultural engineering. 2026. 172 p. Tese (Doutorado em Engenharia Agrícola) - Universidade Federal de Lavras, Lavras, 2026.
dc.identifier.uri.fl_str_mv https://repositorio.ufla.br/handle/1/61086
identifier_str_mv BRANDÃO JÚNIOR, Luiz Carlos. A methodology for synthetic data generation and validation applied to agricultural engineering. 2026. 172 p. Tese (Doutorado em Engenharia Agrícola) - Universidade Federal de Lavras, Lavras, 2026.
url https://repositorio.ufla.br/handle/1/61086
dc.language.iso.fl_str_mv eng
language eng
dc.rights.driver.fl_str_mv Attribution 3.0 Brazil
http://creativecommons.org/licenses/by/3.0/br/
info:eu-repo/semantics/openAccess
rights_invalid_str_mv Attribution 3.0 Brazil
http://creativecommons.org/licenses/by/3.0/br/
eu_rights_str_mv openAccess
dc.publisher.none.fl_str_mv Universidade Federal de Lavras
Escola de Engenharia (EENG)
dc.publisher.program.fl_str_mv Programa de Pós-Graduação em Engenharia Agrícola
dc.publisher.initials.fl_str_mv UFLA
dc.publisher.country.fl_str_mv brasil
publisher.none.fl_str_mv Universidade Federal de Lavras
Escola de Engenharia (EENG)
dc.source.none.fl_str_mv reponame:Repositório Institucional da UFLA
instname:Universidade Federal de Lavras (UFLA)
instacron:UFLA
instname_str Universidade Federal de Lavras (UFLA)
instacron_str UFLA
institution UFLA
reponame_str Repositório Institucional da UFLA
collection Repositório Institucional da UFLA
bitstream.url.fl_str_mv https://repositorio.ufla.br/bitstreams/47d6fd07-6439-4e54-97ae-77a644f1b384/download
https://repositorio.ufla.br/bitstreams/af253796-395e-4b46-9117-c1a0717842b7/download
https://repositorio.ufla.br/bitstreams/6d64f8fa-7bc2-4955-9be5-39e327139fda/download
https://repositorio.ufla.br/bitstreams/403d9f45-e622-4419-878d-edc1a7cf8703/download
https://repositorio.ufla.br/bitstreams/0e649d04-d9a8-4f26-8ee8-601fbc8e33e0/download
https://repositorio.ufla.br/bitstreams/af65f522-019d-4383-b711-b2da5feec9d2/download
https://repositorio.ufla.br/bitstreams/6325e64f-e4bf-41f1-a2d8-a9823698e6b6/download
https://repositorio.ufla.br/bitstreams/bdb52171-2f84-4563-a4ed-549dc91604e0/download
bitstream.checksum.fl_str_mv 5a033ee506f3a0a175bee8fc81f0bd66
dc1a173fe9489e283d3a1f54f6ab2ab9
23e5b2419024f37919fe36ecce4eebf7
d13b9ee9893f27a614510016898db2a7
c85082449256160eff49955d38f1f236
fe3206374123db5d1ff9bf27133cc2b0
af20e4165ca68393f1235565f39a7967
361f7d26001ffad8bb602dc4a509c70d
bitstream.checksumAlgorithm.fl_str_mv MD5
MD5
MD5
MD5
MD5
MD5
MD5
MD5
repository.name.fl_str_mv Repositório Institucional da UFLA - Universidade Federal de Lavras (UFLA)
repository.mail.fl_str_mv nivaldo@ufla.br || repositorio.biblioteca@ufla.br
_version_ 1879593867187060736