Melhorando a tomada de decisões na construção: Modelagem não paramétrica de atrasos induzidos pelo clima

• Ano de defesa: 2024
• Autor(a) principal: Comito, Mateus Borges
• Orientador(a): Izbicki, Rafael
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa Interinstitucional de Pós-Graduação em Estatística - PIPGEs
• Palavras-chave em Português: Clima; Seleção de modelo; Estimação de distribuição; Monte Carlo
• Palavras-chave em Inglês: Weather; Model selection; Distribution estimation
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA::ESTATISTICA::ANALISE DE DADOS
• Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/20146

Effective construction project management faces significant challenges due to frequent delays, many of which are influenced by climatic variables. Anticipating these delays is crucial, and although various methods based on stochastic generators, productivity impact models, or machine learning exist, there is still a notable lack of direct approaches that model productivity using exclusively historical weather data, which is easily accessible. Moreover, it is not sufficient to have only a point estimate of the delay for a specific project. It is more useful to estimate the total uncertainty associated with this estimate. This dissertation proposes a flexible non-parametric model aimed at filling these gaps by estimating the probability distribution of a project's execution time using only weather information. We start from the premise that each task has a daily probability of execution. This process involves a non-stationary stochastic process, described by a non-stationary discrete-time Markov Chain. We use exclusively climatic data to calculate the necessary parameters, and the distribution is estimated through Monte Carlo simulation. The results highlight the utility of the model in predicting optimal start dates, accurately estimating project completion, establishing contractual limits for expected delays due to weather conditions, and analyzing critical paths within the project. Additionally, we present a mathematically rigorous tool for model comparison, allowing for the optimization of hyperparameters and the selection of the most suitable prediction model. This investigation contributes to improving decision-making, minimizing the negative impacts of uncertainty on productivity and construction timelines, resulting in an overall improvement in project efficiency.