Brangus Selection: algoritmo de otimização para seleção de acasalamentos de bovinos com foco na maximização dos ganhos financeiros

Abstract

The growth of the world population, estimated at 9.7 billion by 2050, will determine a greater demand for food, implying new challenges for agriculture and livestock, in the sense of producing more with fewer resources in a scenario where the availability of water and land will be increasingly scarce for these activities. Currently, Brazil has one of the largest bovine herds globally, being one of the leading producers and exporters of beef, together with the USA and the European Union. One way to meet the country’s demand for meat and maintain competitiveness in the international market is to improve the genetic potential of the herds to obtain better production results. Animal breeding consists of selective procedures whose objective is to improve the next generations’ characteristics continuously. This work presents an optimization algorithm based on a selection index to determine an ideal mating system for Brangus cattle. Through a single value, a selection index expresses the total genetic merit of each animal. In other words, it determines the animal’s contribution to the herd considering a specific set of characteristics in a weighted manner that allows the achievement of selection objectives, which are the ultimate purpose of the genetic improvement that will bring the financial return to the production system. The search for the best combination of matings given a selection index is an optimization problem. One of this dissertation’s goals is to investigate the complexity of the mating selection problem in search of an optimal algorithm with a polynomial execution time in the number of animals involved. Also, this work differs by optimizing matings using a selection index based on economic values. Economic values make it possible to assess the economic importance of animal characteristics in a productive system, expressing in monetary values – the financial return that results from modifying a feature through genetic improvement. The proposed solution relies on linear programming and branch-and-bound techniques as a combined way of solving the original problem, modeled as an integer programming instance. The results show that the algorithm provides the optimal solution to the problem within a polynomial-time limit.