Abstract for: Integration of System Dynamics and Machine Learning: A Comprehensive Literature Review

This study synthesizes 30 studies (2001-2026) on integrating System Dynamics and Machine Learning for complex systems modeling. SD provides causal structure for policy analysis, while ML enables data-driven prediction. Four integration approaches are identified: parameter estimation, policy optimization, hybrid frameworks, and structure learning. Applications span health, supply chains, energy, and sustainability. Key gaps remain in theoretical frameworks, scalability, validation, and standardized tools. The literature identifies four primary SD-ML integration approaches, each with distinct advantages and limitations. Quantitative analysis reveals a strong preference for Hybrid Simulation Frameworks, where practitioners maintain SD and ML as interacting yet distinct modules. Meanwhile, Data-Driven Structure Learning remains a niche but promising frontier. The following subsections detail the mechanics, benefits, and constraints of each methodology. The study concludes that promising directions include three-way integration of SD, ML, and agent-based modeling, and combining SD-ML with optimization, control theory, or discrete-event simulation for operational contexts. Additionally, developing flexible multi-paradigm frameworks and establishing theoretical foundations for integrating diverse modeling approaches based on specific objectives emerge as key outcomes for advancing hybrid system modeling. High-impact research directions include developing theoretical frameworks, robust validation methodologies, bidirectional integration methods, and open-source tools. SD-ML integration promises to address complex societal challenges by combining causal structure with predictive power. Realizing this potential requires sustained research to overcome theoretical and practical challenges, enabling transformative contributions to complex system analysis. Artificial intelligence was used for paraphrasing and grammar checking.