Most of the ruminant production in Brazil is based on the use of pastures, which are present in every Brazilian biome. The aim of this study was to extract patterns and knowledge from the stakeholders’ response database regarding the technological processes used for pasture conservation and management. Electronic questionnaires were used to perform the stakeholders’ survey, as they are more economical and agile. 712 people from all Brazilian regions and biomes were interviewed between July and August 2019. Four technologies were selected by pasture specialists from Embrapa Pecuária Sudeste to be analyzed as dependent variables. The techniques were classified as adopted (1) or not adopted (0). The generalized linear model (GLM) method was used to estimate statistical parameters, and the analysis was directed to the evaluation of the combination of statistically significant independent technological processes (P < 0.05). The database was then submitted to an artificial mining process, with the application of the decision tree induction method, which involves hierarchical models used for their predictive capacity. The combined use of parametric and non-parametric methodologies effectively facilitated the extraction of patterns, knowledge, and insights into stakeholders’ decision-making processes regarding the integration of technological practices in structuring pasture conservation and production systems.

Anais do 61º Congresso da Sociedade Brasileira de Economia, Administração e Sociologia Rural (SOBER) (978-85-5722-931-0) - Padrões De Produção Da Bovinocultura De Corte No Bioma Cerrado Segundo O Censo Agropecuário 2017

O objetivo deste trabalho foi, por meio de técnicas de mineração de dados, classificar os animais de dois rebanhos da Embrapa, em função das réguas de DEP genômicas (DEPg), para identificar os principais atributos (características) que direcionam o entendimento dos diferentes objetivos de seleção nos dois rebanhos. Selecionaram-se oito atributos, para análise de classificação dos animais. Para realizar a classificação dos animais em função do rebanho foram utilizados três algoritmos supervisionados, buscando verificar qual apresentaria o melhor desempenho: árvore de decisão (J48), árvores de modelo logístico (LMT) e floresta randômica (Random Forest - RF). O algoritmo mais acurado foi o Random Forest, que modelou os dados com maior ajuste e acurácia.