The study titled “Ensaios em matemática aplicada: estimação e trajetórias bootstrap de oferta de sangue e estudo de desempenho de extensões do algoritmo de Programação Dinâmica Dual Estocástica” [Essays in applied mathematics: blood supply estimating and bootstrap trajectories and performance study on the extensions of the Stochastic Dual Dynamic Programming algorithm] won the Graduate Award of the Brazilian Society of Applied and Computational Mathematics (SBMAC), in the Master’s category. The work carried out by student Michelle Bandarra, and advised by FGV’s School of Applied Mathematics (EMAp) professor Vincent Gérard Guigues, analyzes the estimation and generation of future series trajectories of blood supply and studies the different extensions of the Stochastic Dual Dynamic Programming (SDDP) algorithm.

The dissertation calibrated Auto Regressive Vector (VAR) models to generate supply trajectories of Hemorio blood bags using two different bootstrap techniques, based on the non-normality of the model’s errors. The study concluded that both techniques are suitable approaches to improve predictability in blood supply series. The second topic of the work was a comparison, under the optics of portfolio selection models, between the computational performance of two algorithms. The first is a modification of the SDDP that calculates multiple cuts per iteration – the Multicut Decomposition Algorithm (MuDA). The second introduces cut selection strategies to MuDA, which the work calls Cut Selection Multicut Decomposition Algorithm (CuSMuDA) – which is unprecedented in academic literature.

The SBMAC graduate award was first granted in the 25^th CNMAC, held in 2002. Since then, the award in the PhD and Master’s categories has been awarded annually to the best PhD thesis and Master’s dissertation submitted to the Judging Panel.

The awarded Master’s dissertation is available on FGV’s Digital Library.