Análise comparativa de eficiência na pecuária leiteira: identificando as possibilidades de sucesso

Abstract

Milk is one of the most consumed products in the world. In addition to being essential in the early stages of human life, milk also has a strong presence in the economic context. Consequently, many families of small and medium producers depend on its production for income generation. In Brazil, one of the world's largest producers, production is concentrated in the states of Minas Gerais and Paraná. Factors such as proper management, good production management practices, and the introduction of technologies have been responsible for raising the quality and productivity of milk, marking a new phase in dairy farming in the country. However, the climate has been a worsening factor for the sector, due to global temperature changes and extended periods of drought. This is because the animals depend on adequate temperatures for production, and their feed relies on grain production, which is compromised by the lack of rain. Thus, high production costs and a decrease in production quantity are observed. In this context, this study aimed to evaluate the performance of milk-producing regions in Brazil in 2022 using Principal Component Analysis (PCA), Data Envelopment Analysis (DEA), Monte Carlo Simulation (MCS), and Artificial Neural Networks (ANN). To this end, the study followed the standards of descriptive research, with a quantitative approach and inductive logic. The study was conducted over 24 months, using the software R-Project 3.2.2 and RStudio 2023.12.0+369 to support the application of the techniques. The results of this study showed that only 35% of producers achieved maximum pure technical efficiency, with about 91% of the total number of producers having production areas smaller than 100 hectares. The probabilistic stage provided valuable insights from the best adjustments of the model variables based on the Log-Logistic, Pearson, and Log-Normal functions. The results of the ANN technique application indicated good performance of the neural network for the monthly classification of milk price dynamics, with an accuracy of 87.7%, precision of 86.57%, and a mean squared error (MSE) of 0.1229. It is expected that this study will provide a solid knowledge base for the sector's stakeholders, as well as support the decision-making of producers and/or managers on their properties.