Modeling Discrete-Event Simulations Using Natural Language Processing: A Healthcare Application

Abstract

Discrete-event simulation (DES) is widely used to model complex healthcare systems; however, manually developing these simulation models often requires extensive effort and specialized expertise. This study explored how natural language processing (NLP) techniques can automate DES model generation from text descriptions and optimize resource allocation in the healthcare domain. We used GPT-4o large language model to demonstrate that DES models can be automatically generated from natural language prompts with accuracy comparable to traditional simulation software. The GPT-4o model successfully simulated a skin care clinic and a complex medical care facility, producing results aligned with Arena software for metrics such as average queue times and patient throughput. Additionally, GPT-4o determined the optimal resource allocation to minimize costs while satisfying the patient waiting time constraints. The automated generation of simulations shows the potential to combine NLP with DES to accelerate healthcare system modeling and optimization.