Estudo e aprendizado de redes neurais artificiais para aplicações diversas
| dc.contributor.advisor | Lotufo, Anna Diva Plasencia [UNESP] | |
| dc.contributor.author | Campagnolo, Maria Fernanda Lenquist [UNESP] | |
| dc.contributor.committeeMember | Lotufo, Anna Diva Plasencia [UNESP] | |
| dc.contributor.committeeMember | Chaucanes, Paula Andrea Montenegro | |
| dc.contributor.committeeMember | Formaggio, Giovanni Panegossi | |
| dc.contributor.institution | Universidade Estadual Paulista (Unesp) | pt |
| dc.date.accessioned | 2026-01-13T16:53:39Z | |
| dc.date.issued | 2025-12-15 | |
| dc.description.abstract | Analisa-se a demanda de viagens aéreas no Brasil, com foco na variável Horas Voadas, por meio de um sistema de previsão inteligente. O estudo visa modelar a complexi- dade e a não-linearidade inerentes à série temporal, incorporando o impacto de choques estruturais, como a pandemia de COVID-19. Desenvolve-se uma arquitetura de Redes Neurais Artificiais (RNA) do tipo Multilayer Perceptron (MLP) utilizando o algoritmo de otimização de Regularização Bayesiana (trainbr) no ambiente MATLAB. A base de dados, referente ao período de 2010 a 2025, foi rigorosamente dividida em 70% para o treinamento e 30% para as etapas de validação e teste, assegurando a imparcialidade na avaliação. A topologia final do modelo, com 24 neurônios na camada oculta, otimiza a capacidade de generalização. O modelo demonstrou alta fidelidade e consistência, ates- tada pelo Coeficiente de Correlação (R) geral de 0.9855 e pelo R2 de 0.7803 no conjunto de Teste. A análise de resiliência durante a crise (2020-2021) mantém o Erro Percentual Absoluto Médio (MAPE) de 1,67%. Conclui-se que o sistema estabelece uma ferramenta robusta para o planejamento estratégico e a mitigação de riscos, ressalvando-se que pro- jeções para anos subsequentes a 2025 são consideradas inconclusivas devido à ausência de dados observados e à incerteza estrutural do setor. | pt |
| dc.description.abstract | Analysis of air travel demand in Brazil, focusing on the variable Flight Hours, is conducted through an intelligent forecasting system. The study aims to model the complexity and non-linearity inherent to the time series, incorporating the impact of structural shocks, such as the COVID-19 pandemic. An Artificial Neural Networks (ANN) architecture of the Multilayer Perceptron (MLP) type is developed, utilizing the Bayesian Regularization (trainbr) optimization algorithm within the MATLAB environment. The database, spanning the period from 2010 to 2025, was strictly divided into 70% for training and 30% for validation and testing, ensuring impartiality in the assessment. The model’s final topology, featuring 24 hidden neurons, optimizes generalization capability. The model demonstrated high fidelity and consistency, attested by the overall Coefficient of Correlation (R) of 0.9855 and an R2 of 0.7803 on the Test set. Resilience analysis during the crisis (2020-2021) maintained a Mean Absolute Percentage Error (MAPE) of just 1.67%. It is concluded that the system establishes a robust and essential tool for strategic planning and risk mitigation within the Brazilian aeronautical sector, with the caveat that projections for years subsequent to 2025 are deemed inconclusive due to the lack of observed data and inherent structural uncertainty. Keywords: Artificial Neural Networks; Time Series Forecasting; Aviation; Machine Learning; Bayesian Regularization; MATLAB. | en |
| dc.description.sponsorshipId | Não há | pt |
| dc.identifier.citation | CAMPAGNOLO, Maria Fernanda Lenquist. Estudo e aprendizado de redes neurais artificiais para aplicações diversas. Orientadora: Anna Diva Plasencia Lotufo. 2025. Trabalho de conclusão de curso (Graduação em Engenharia Elétrica) - Universidade Estadual Paulista (UNESP), Faculdade de Engenharia, Ilha Solteira, 2025 | pt |
| dc.identifier.lattes | 0815781368310607 | |
| dc.identifier.orcid | 0009-0000-8954-110X | |
| dc.identifier.uri | https://hdl.handle.net/11449/318353 | |
| dc.language.iso | por | |
| dc.publisher | Universidade Estadual Paulista (Unesp) | pt |
| dc.rights.accessRights | Acesso aberto | pt |
| dc.subject | Redes Neurais Artificiais | pt |
| dc.subject | Previsão de Séries Temporais | pt |
| dc.subject | Regularização Bayesiana | pt |
| dc.subject | Machine Learning | en |
| dc.subject | MATLAB | en |
| dc.title | Estudo e aprendizado de redes neurais artificiais para aplicações diversas | pt |
| dc.title.alternative | Study and learning of artificial neural networks for diverse applications | en |
| dc.type | Trabalho de conclusão de curso | pt |
| dspace.entity.type | Publication | pt |
| relation.isAuthorOfPublication | eaa328dd-5196-4315-a604-23b05d79e89e | |
| relation.isAuthorOfPublication.latestForDiscovery | eaa328dd-5196-4315-a604-23b05d79e89e | |
| relation.isOrgUnitOfPublication | 85b724f4-c5d4-4984-9caf-8f0f0d076a19 | |
| relation.isOrgUnitOfPublication.latestForDiscovery | 85b724f4-c5d4-4984-9caf-8f0f0d076a19 | |
| unesp.campus | Universidade Estadual Paulista (UNESP), Faculdade de Engenharia, Ilha Solteira | pt |
| unesp.examinationboard.type | Banca pública | pt |
| unesp.undergraduate | Ilha Solteira - FEIS - Engenharia Elétrica | pt |
Arquivos
Pacote original
1 - 1 de 1
Carregando...
- Nome:
- campagnolo_mfl_tcc_ilha.pdf
- Tamanho:
- 2.85 MB
- Formato:
- Adobe Portable Document Format
Licença do pacote
1 - 2 de 2
Carregando...
- Nome:
- license.txt
- Tamanho:
- 2.14 KB
- Formato:
- Item-specific license agreed upon to submission
- Descrição:
Carregando...
- Nome:
- campagnolo_mfl_autorizacao_ilha.pdf
- Tamanho:
- 64.8 KB
- Formato:
- Adobe Portable Document Format
- Descrição:

