Analytical, Simulation, and Learning-Based Approaches for Decision Support in Stochastic Service Systems: A Comparative Study

Abstract

This paper presents a comparative analysis of three major methodologies for modeling stochastic service systems: analytical modeling (AM), simulation modeling (SM), and learning-based modeling (LM). While these approaches have been widely studied in isolation, relatively little research has systematically examined their respective roles and complementarities in supporting decision making for complex service systems. Using representative applications from international border-crossing operations and healthcare service systems, this study evaluates the strengths and limitations of the three methodologies in terms of modeling assumptions, interpretability, computational requirements, and predictive capability. The results show that analytical models provide valuable structural insights and prescriptive guidance under simplifying assumptions. Simulation models offer flexible and realistic representations of system dynamics for policy evaluation, and learning-based models deliver strong predictive performance in data-rich and highly dynamic environments. In addition to comparing these approaches, the paper proposes an integrated decision-support framework in which learning-based models generate short-term forecasts, analytical models design operational policies, and simulation models evaluate policy performance under realistic conditions. The findings highlight the complementary nature of these methodologies and provide practical guidance for selecting and integrating modeling approaches to support decision making in stochastic service systems.