For over 40 years, there have been significant advancements in the theory, methods, and application of risk analysis (Greenberg et al., 2020). The focus of the field has been on answering questions ranging from the identification of what can go wrong to how key stakeholders can be informed of the outcomes of risk analysis for purposes of decision making (Greenberg et al., 2012). A key challenge analysts and decisionmakers face is the treatment of uncertainty in risk models. This challenge has remained at the forefront of risk science and can be traced back to the early efforts of risk quantification (Kaplan & Garrick, 1981). Risk analysts have thus employed progressively more powerful methods over time to estimate and explain the magnitude of risk and uncertainty. There has been much intellectual investment into predictive modeling to forecast and measure risk more precisely. The field has benefited from technological developments that have made significant amounts of cross-sectional and longitudinal data available for risk analysis. Increased data availability has also made increasingly sophisticated risk assessments possible (Choi & Lambert, 2017; Greenberg et al., 2020). However, despite having large amounts of data available, the lack of precision in characterizing the size, usefulness, and robustness of data remain (Nateghi & Aven, 2021). Despite significant advancements in the long history of risk science (Aven & Flage, 2020), uncertainty remains in the complex and dynamic risk environment. Risk analysts continue to grapple with issues of incorporating uncertainty in dynamic risk models, while key stakeholders constantly face the pressures of decision making under evolving and highly challenging conditions. This special issue explores the employment of Bayesian networks (BNs, also called Bayes nets or Bayesian belief networks) as a versatile and powerful framework to model complex systems, (e.g., Pourret et al., 2008) and for reasoning and decision making under uncertainty (Jensen, 1996).
Paper
Full text
Bayesian networks for risk analysis and decision support
Semantic Scholar · Computer Science · 2022
Abstract
For over 40 years, there have been significant advancements in the theory, methods, and application of risk analysis (Greenberg et al., 2020). The focus of the field has been on answering questions ranging from the identification of what can go wrong to how key stakeholders can be informed of the outcomes of risk analysis for purposes of decision making (Greenberg et al., 2012). A key challenge analysts and decisionmakers face is the treatment of uncertainty in risk models. This challenge has remained at the forefront of risk science and can be traced back to the early efforts of risk quantification (Kaplan & Garrick, 1981). Risk analysts have thus employed progressively more powerful methods over time to estimate and explain the magnitude of risk and uncertainty. There has been much intellectual investment into predictive modeling to forecast and measure risk more precisely. The field has benefited from technological developments that have made significant amounts of cross-sectional and longitudinal data available for risk analysis. Increased data availability has also made increasingly sophisticated risk assessments possible (Choi & Lambert, 2017; Greenberg et al., 2020). However, despite having large amounts of data available, the lack of precision in characterizing the size, usefulness, and robustness of data remain (Nateghi & Aven, 2021). Despite significant advancements in the long history of risk science (Aven & Flage, 2020), uncertainty remains in the complex and dynamic risk environment. Risk analysts continue to grapple with issues of incorporating uncertainty in dynamic risk models, while key stakeholders constantly face the pressures of decision making under evolving and highly challenging conditions. This special issue explores the employment of Bayesian networks (BNs, also called Bayes nets or Bayesian belief networks) as a versatile and powerful framework to model complex systems, (e.g., Pourret et al., 2008) and for reasoning and decision making under uncertainty (Jensen, 1996).