European Journal of Operational Research ( IF 6.4 ) Pub Date : 2022-11-23 , DOI: 10.1016/j.ejor.2022.11.038 Zhenxia Cheng, Jun Luo, Ruijing Wu
We consider the simulation optimization problem of selecting the best system design from a finite set of alternatives, which is known as ranking and selection (R&S). Many fully sequential procedures have been proposed to solve the R&S problem using a static sampling rule in order to ensure a finite-sample statistical guarantee. In this paper, we develop fully sequential procedures that can incorporate various adaptive sampling rules, based on a modification of Paulson’s bound (Paulson, 1964), while still preserving the finite-sample guarantee. In particular, we propose an adaptive sampling rule that utilizes the consecutively updated sample mean and sample variance information by solving a minimization problem of the approximated total sample size. Finally, we demonstrate the efficiency of the proposed procedures with several existing procedures through extensive simulation experiments, and apply them to solve an ambulance dispatching problem.
中文翻译:
关于自适应完全顺序程序的有限样本统计有效性
我们考虑从有限的备选方案中选择最佳系统设计的仿真优化问题,这被称为排序和选择 (R&S)。为了确保有限样本统计保证,已经提出了许多完全顺序的程序来使用静态采样规则来解决 R&S 问题。在本文中,我们开发了完全顺序的程序,可以结合各种自适应采样规则,基于保尔森界限的修改(保尔森,1964 年),同时仍然保留有限样本保证。特别是,我们提出了一种自适应采样规则,该规则通过解决近似总样本量的最小化问题来利用连续更新的样本均值和样本方差信息。最后,