Otimização da Seleção de Acasalamentos para Melhoramento Genético Animal

Abstract

Demographic changes in the world may increase the demand for livestock products. Thus, strategies that contribute to a greater offer of these products, and that add quality and sus tainability, need to be formulated. PampaPlus is a breeding program that aims to analyze the genetic performance of cattle and provide tools that allow the direction of breeding work. An important practice in this type of program is the selection of matings. In this sense, a genetic algorithm was previously proposed to analyze phenotypic data of the selected animals and recommend matings to producers. Experiments performed proved its effectiveness. However, the approach needs improvement, mainly from performance. The algorithm used a stopping criterion based on a maximum number of iterations, but generally reached convergence for an answer to the problem before the pre-established limit. This made it run longer and with iterations that did not provide any significant im provement in genetic indexes. In this sense, two approaches to stopping criteria found in the literature were implemented, and test cases were created for validation and compari son, each with a set of animals. The first approach could not be validated in different test cases. However, the second approach was validated and presented performance results with a reduction of up to 50.8% in processing time compared to the original stopping cri terion, maintaining the same genetic qualification indexes achieved. Improvements were also implemented in the restriction of the use of bulls, because, in some situations, the algorithm presented recommendations that exceeded a limit. After implementation, the results indicated that the mating recommendation given as an answer by the algorithm obeyed the usage restriction imposed. In addition, some routines and data structures have been refactored to eliminate dependencies with genetic characteristics measured and used in PampaPlus, thus making it possible to use the algorithm in other breeding programs. After implementing improvements, a benchmark performed showed that the genetic al gorithm obtains mating recommendations with better genetic qualification indexes when compared to another mating recommendation tool.