Árvores hibridas: particionamento recursivo baseado em modelos paramétricos

• Autor(a) principal: Santiago, Vinicius Hideki Yamada
• Data de Publicação: 2022
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFSCAR
• Texto Completo: https://repositorio.ufscar.br/handle/ufscar/16050
• topic: Árvore híbrida; Árvore de decisão; Modelo linear generalizado; Hybrid tree; Decision tree; Generalized linear model; CART; Regression tree; CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA::ESTATISTICA; CIENCIAS AGRARIAS::ZOOTECNIA::PRODUCAO ANIMAL

The present study seeks to explore hybrid trees, so called called recursive partitioning based on parametric models. Such method is based on traditional decision trees (regression and classification tree) and parametric models (generalized linear models). With this, it is expected to explain the functioning of hybrid trees, both in theoretical aspects and computational implementations. Finally, the hybrid trees technique will be applied to a real problem concerning pre-slaughter losses of broilers, stating the obtained results, in order to highlight the advantages and disadvantages of this technique. As a conclusion, it was noticed that model-based recursive partitioning is a technique best used if applied to small to medium-sized databases.