Approximate Simulation Budget Allocation for Subset Ranking

Accurate performance evaluation of discrete event systems needs a huge number of simulation replications and is thus time-consuming and costly. Hence, efficiency is always a big concern when simulations are conducted. To drastically reduce its cost when conducting them, ordinal optimization emerges. To further enhance the efficiency of ordinal optimization, optimal computing budget allocation (OCBA) is proposed to decide the best design accurately and quickly. Its variants have been introduced to achieve goals with distinct assumptions, such as to identify the optimal subset of designs. They are restricted in selecting the best design or optimal subset of designs. However, a highly challenging issue, i.e., subset ranking, remains unaddressed. It goes beyond best design and optimal subset problems. This work develops a new OCBA-based approach to address the issue and establishes its theoretical foundation. The numerical testing results show that, with proper parameters, it can indeed enhance the simulation efficiency and outperform other existing methods in terms of the probability of correct subset ranking and computational efficiency.