Lógica fuzzy na seleção simultânea em couve e bata-doce

• Ano de defesa: 2019
• Autor(a) principal: Ana Clara Goncalves Fernandes
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Inteligência computacional; Brassica oleracea L var; Acephala DC; Ipomoea batatas (L) LAM; Índices de seleção; Batata-doce; Inteligência artificial; Couve; Lógica difusa
• Link de acesso: http://hdl.handle.net/1843/ICAS-BBLFVF

The genetic improvement aims at obtaining agronomically superior plants to the preexisting ones. For this, techniques are used for the simultaneous selection of characters. However, there are limitations intraditional procedures regarding the selection of characteristics for desired commercial ranges and qualitative characteristics. In this context, fuzzy logic can enable the computational modeling of the researcher's experience considering qualitative characters and with predefined intervals. Therefore, the objective of this work was to develop a fuzzy logic system for the simultaneous selection of several characteristics in the genetic improvement of cabbage and sweet potato and to test its efficiency by comparing it with the Mulamba and Mock index. For this, data from two experiments were used: (1) randomized complete block design (DBC) with 24 families of half-siblings of cabbage, using four replicatesand five plants per plot; (2) 24 sweet potato genotypes in DBC with four replicates and 10 plants per plot. In the experiment 1 were evaluated quantitative characteristics associated with the ease of culturaltreatments and the productivity of leaves and qualitative characteristics associated with leaf quality. In the experiment 2 the following were evaluated: productivity of dry mass of branches, productivity of commercialroot, commercial root mean weight, root shape, insect resistance, root dry matter content and the productivity of root scrap. For all quantitative characters the genetic values were obtained via REML / BLUP.Subsequently, the selection gain was estimated by the Mulamba and Mock index. The fuzzy systems were developed using software R, being used the Mandani Min methodology in the fuzzification stage and thecentroid method in the defuzzification. The fuzzy logic was efficient in simultaneous selection, providing gains close to those obtained by the Mulamba and Mock method for half-siblings of cabbage, besidesadditionally allowing the selection of qualitative characters. For the experiment with sweet potato, the fuzzy logic allowed gains similar to those obtained by the Mulamba and Mock methodology for the FIM system, however, the fuzzy system allowed a greater gain for the commercial roots productivity, which is the most important characteristic for the improvement. Already for the characteristics related to the AH and AA systems the gains were higher than those obtained by the index of MM. The fuzzy systems allowed the selection to commercially predefined intervals, proving to be a promising tool for genetic improvement.