Taxonomy, Tools, And A Framework For Combining Simulation Models With AI/ML Models

There exist compelling use cases that demonstrate a combination of discrete event simulation and AI/ML approaches is not only possible but may yield improved results, despite their differences. This is illustrated well in the context of healthcare, particularly in an emergency department, where hard resource constraints and vague decision-making from large datasets, means neither approach sufficiently captures the entire picture of patient care. A hybrid approach may see simulation models providing details of paths that patients follow through processing, which covers associated resource utilization, and AI/ML models providing deeper insight into the treatment details of patients and their associated experiences. This paper provides a taxonomy and architecture for combining simulation models and AI/ML models. In addition, it details a concrete implementation with a Colored Petri Nets (CPN)-based simulation models coupled with Python-based AI/ML models. We illustrate the framework with several examples from the healthcare domain.