Modelagem de um adaptive neuro fuzzy inference system para análise de risco em projetos.

Detalhes bibliográficos
Ano de defesa: 2016
Autor(a) principal: França, Daniel cruz de
Orientador(a): Não Informado pela instituição
Banca de defesa: Não Informado pela instituição
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Universidade Federal da Paraíba
Brasil
Engenharia de Produção
Programa de Pós-Graduação em Engenharia de Produção
UFPB
Programa de Pós-Graduação: Não Informado pela instituição
Departamento: Não Informado pela instituição
País: Não Informado pela instituição
Palavras-chave em Português:
Link de acesso: https://repositorio.ufpb.br/jspui/handle/tede/8163
Resumo: Several researches highlight the importance of risk management in project management. Many authors propose traditional models with statistical and deterministic methods, though some risk project management issues are based on conceptual frameworks, expert opinion and human experience. This kind of problem makes difficult the use of classical models, but can be mathematically treated using fuzzy logic. In addition, historical data of projects can provide information about the organization's risk analysis experience and be modelled by a learning mechanism. The method used in this work is the Adaptive Neuro-fuzzy Inference System (ANFIS), which is capable of aggregating the mathematical treatment capacity of conceptual models with a hybrid learning algorithm. Thus, the aim of this study is to model an ANFIS that is able to analyze the risks of projects. A set of projects was analyzed by means of a risk management checklist with factors arranged in a risk breakdown structure (RBS). Estimates were made using probability and impact matrix, and expert opinion. The risk of each project was defined as an integer between 1 and 10. To select the best model among 32 different ANFIS settings, 84% of the data were used in 10-fold cross-validation. The model with the best results in validation process was selected and tested with the remaining data. The results attained in the evaluation were: mean squared error (MSE) of 0.2207, mean absolute error (MAE) of 0.3084, coefficient of determination (R²) of 0.9733 and 80% of accuracy. These results indicate that the project risk management can be successfully performed by ANFIS. This enables the modeling of knowledge and human experience and can reduce costs of skilled labor and improve the speed of analysis.