Uso de analytics e redes neurais perceptron para investigação das relações entre insumos agrícolas e produtividade

• Autor(a) principal: Sales, Rafael Francisco Czornik
• Data de Publicação: 2024
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UTFPR (da Universidade Tecnológica Federal do Paraná (RIUT))
• Texto Completo: http://repositorio.utfpr.edu.br/jspui/handle/1/35634
• topic: Inteligência artificial; Agroindústria; Redes neurais (Computação); Artificial intelligence; Agricultural industries; Neural networks (Computer science); CNPQ::ENGENHARIAS; Engenharia/Tecnologia/Gestão

Technological advancements in areas such as advanced sensors, GPS systems, drones, and precision fertilization techniques have been significant drivers of the evolution of agribusiness. This study employed Analytics methods and Perceptron neural networks to investigate correlations between agricultural inputs and the productivity of soybean and corn crops from 2020 to 2023, considering variables such as the use of herbicides, fungicides, fertilizers, and climatic indicators. Big Data analysis enabled the identification of useful patterns and trends, using tools like Python and scientific libraries for data processing and visualization. The Perceptron demonstrated potential as a lightweight and scalable computational model, although the results revealed no explicit correlations between inputs and productivity, suggesting that additional factors, such as climate and soil management, play a significant role. The experiments also highlighted computational limitations that restricted model effectiveness, emphasizing the need for more robust hardware and supplementary data to improve analysis accuracy. Despite these challenges, the study underscores the value of using advanced technologies in rural management, providing a solid foundation for strategic and optimized decision-making in the agricultural sector. It concludes that combining Analytics techniques with neural networks holds great potential for future analyses, provided that more external variables and computational resources are integrated into the process, fostering more effective agricultural management adapted to the increasing demands of the agribusiness market.