ABSTRACT The objective of this research was to simulate the genetic gains expected comparing random mating strategies and mate selection by optimum contribution with different penalty levels in the inbreeding rate of Santa Inês sheep. The optimum contribution theory was thus applied to optimize genetic gain in the long term in twelve selection groups by selectively mating 500 females with the respective males, increasingly penalizing the increase in inbreeding in the objective function. Genetic algorithms were used to find the optimum contribution. Optimization was achieved via EVA software. Selection candidates had their contribution defined into four treatments, using different values to weigh the genetic merit and penalize increases in inbreeding. This made it possible to measure the degree of control over those parameters that can be obtained with this methodology. This selection offers different levels of genetic gain, which are achievable from restrictions on the coancestry. The number of males selected and their distribution into selection groups varied according to the penalty attributed to inbreeding in the objective function. Mate selection using optimum contribution should be adopted when aiming to limit the increase in inbreeding. Increasing the exchange of genetic material between groups is recommended to elevate genetic gain and maintain control over inbreeding.
Dentre as várias técnicas de previsão de vazão, os modelos guiados por dados (DDMs: data-driven models) estão sendo muito utilizados. Estes se baseiam num banco de dados formado pelos registros históricos das variáveis de entrada (precipitação e vazão) e saída (vazão) para realizar a previsão. Redes neurais artificiais (ANNs: artificial neural networks) são os tipos de DDMs mais comuns e se mostram normalmente mais precisas do que outros modelos empíricos, mas possuem a desvantagem de não serem suficientemente transparentes. Um dos métodos de aprendizado de máquina que não possui esse problema é o aprendizado baseado em instâncias (IBL: instance-based learning). O modelo k-nearest neighbor (KNN) é um exemplo de IBL. Neste trabalho, variações do modelo KNN são utilizadas e propostas a fim de realizar previsão de vazão em rio do estado de Sergipe. Os resultados são comparados aos de simulações feitas com o uso de redes neurais artificiais e indicam superioridade das ANNs, mas também previsões satisfatórias com o KNN.
Among several streamflow forecasting techniques, data-driven models (DDMs) are widely used. They employ a database formed by historical input (precipitation and streamflow) and output (streamflow) variables to perform the prediction. Artificial neural networks (ANNs) are the most common types of DDMs, and are typically more accurate than other empirical models, but have the disadvantage of not being sufficiently transparent. One of the machine-learning methods that do not have this problem is the so-called "instance‑based learning" (IBL). The KNN algorithm is an example of IBL. This research applies and proposes variations of the KNN model in order to forecast streamflows in a river of the state of Sergipe, Brazil. The results are compared to simulations carried out by the use of artificial neural networks and indicate the superiority of the ANNs, but also show satisfactory forecasts of the KNN.