From divergent paradigms to integrated risk intelligence: synthesizing data- and knowledge-driven approaches in complex projects

Complex engineering projects are increasingly exposed to dynamic, cascading, and interdependent risks. Risk analysis in these projects has evolved, diverging into two dominant paradigms: data-driven approaches, which conceptualize risk as latent statistical patterns extracted from empirical data, and knowledge-driven approaches, which encode heuristics, expert judgment, and institutional knowledge. To critically examine these paradigms, this study conducts a systematic review, classifying modeling approaches into six categories and systematically analyzing their foundational assumptions, technical modeling characteristics, and application contexts. Through this synthesis, it is observed that data-driven approaches offer scalability and predictive power, yet key challenges include interpretability and the lack of standardized benchmark datasets. Knowledge-driven approaches, on the other hand, provide transparency and domain alignment, yet challenges remain in their adaptability, transferability and heavy reliance on experts. Despite their differences, both paradigms commonly rely on a static formulation of risk, which does not capture dynamic feedback, escalation, and evolving interdependencies. In response, the review outlines a future research agenda that emphasizes evidence-support structures (e.g. , knowledge graphs ), dynamic project representations (e.g. , digital twins with retrieval-augmented reasoning ), and temporal-systemic formulations (e.g. , dynamical systems ). Advancing these directions can enable the development of next-generation risk frameworks that are more adaptive, explainable, and transferable across complex project environments.