Aplicação de algoritmo bio inspirado para refinamento de recomendação de adubação para área específica

• Ano de defesa: 2018
• Autor(a) principal: Tessele, Vinícius
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Medianeira; Brasil; Programa de Pós-Graduação em Tecnologias Computacionais para o Agronegócio; UTFPR
• Palavras-chave em Português: Mineração de dados (Computação); Cultivos agrícolas; Software - Desenvolvimento; Data mining; Crops; Computer software - Development; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO; Engenharia/Tecnologia/Gestão
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/3451

Soil management is of fundamental importance for increasing productivity. Among several existing techniques, nutrient replacement is highlighted by efficiency. For this, it is necessary to know by means of soil analysis the deficiencies present in the soil, in order to restore the nutritional needs for the cultivars of that area. Computational technology and the branch of Artificial Intelligence (AI), already present in rural areas, contribute to increase productivity and profitability in agriculture, with several embedded devices and management software. For this study, the technique derived from AI known as Artificial Immunological Systems (SIA) was used. Among its several applications, the one used in this work is the optimization using the aiNet algorithm, for the refinement of fertilization of the studied property. The application of this algorithm aims to verify its adherence and to present the recommendation of fertilization with greater financial and productive advantage. For this, experiments were carried out with lettuce cultivation, which was chosen because it is a fast crop, thus allowing a greater number of replicates of the algorithm. The algorithm was implemented using the object pascal language and the development platform used was the Delphi IDE in the XE10 version that enables the creation of multi-platform applications. 80 seedlings with the initial population and 12 control/control beds were established for the experiment, all with 12 plants per plot, totaling 1104 feet of lettuce. However, for data entry, only four central feet were selected from each of the 80 beds that were weighed, thus obtaining the value of their crude fresh mass and in a second weighing the obtaining of their commercial fresh mass. Seven iterations were performed, which for each iteration the main objective of the application was the optimization of the fertilization result in order to establish the best results. However, the results obtained showed a significant improvement in the performance in the last 3 rounds, with the increase in the weight values of the fresh mass, as well as the reduction of the amount of fertilizer used, the consequence is the obtaining of greater profit in the production. Another positive aspect was the reduction of the standard deviation presented in the rounds of 5.6 and 7. The lower dispersion of the data suggests that the algorithm is being efficient in the optimization and search of the refinement of the fertilization.