A framework proposal for machine learning-driven agent-based models through a case study analysis

Agent-based modeling (ABM) has been widely employed by researchers in various domains. Developing valid and useful agent-based models (ABMs) imposes challenges on the modelers. Using machine learning (ML) techniques in ABMs may facilitate the development of these models and improve their performance. This paper provides a detailed overview of the relationship between ML and ABM approaches. The benefits and drawbacks of data-driven ABMs are evaluated. A main scheme for utilizing ML techniques in ABMs is provided and explored through references to the relevant studies. As part of the primary scheme, a framework for modeling agent behaviors in ABMs utilizing ML approaches is proposed. In the framework, theoretical support is also combined with ML approaches in order to increase the accuracy of agent behavior generated by ML approaches. Using the suggested framework, a real-world case study is performed to investigate the application of ML techniques to improve the accuracy of ABMs and facilitate their creation. The findings indicate that ML approaches may facilitate the construction of ABMs.