Building Bayesian networks based on DEMATEL for multiple criteria decision problems: A supplier selection case study

•We propose a novel and systematic way of building causal decision support models.•Our approach combines probabilistic and multi-criteria decision making tools.•The resulting models are based on knowledge elicited from multiple experts.•The model's consistency with expert knowledge is evaluated...

Full description

Saved in:
Bibliographic Details
Published in:Expert systems with applications 2019-11, Vol.134, p.234-248
Main Authors: Kaya, Rukiye, Yet, Barbaros
Format: Article
Language:English
Subjects:
Bayesian analysis
Bayesian networks
Component and supplier management
Decision analysis
Decision making
Decision support systems
DEMATEL
Multiple criteria decision making
Multiple criterion
Sensitivity analysis
Supplier selection
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•We propose a novel and systematic way of building causal decision support models.•Our approach combines probabilistic and multi-criteria decision making tools.•The resulting models are based on knowledge elicited from multiple experts.•The model's consistency with expert knowledge is evaluated by sensitivity analysis.•The proposed approach is applied to a supplier selection case study. Bayesian Networks (BNs) are effective tools for providing decision support based on expert knowledge in uncertain and complex environments. However, building knowledge-based BNs is still a difficult task that lacks systematic and widely accepted methodologies, especially when knowledge is elicited from multiple experts. We propose a novel method that systematically integrates a widely used Multi Criteria Decision Making (MCDM) approach called Decision Making Trial and Evaluation Laboratory (DEMATEL) in BN construction. Our method elicits causal knowledge from multiple experts based on DEMATEL and transforms it to a BN structure. It then parameterizes the BN by using ranked nodes and evaluates its robustness and consistency by using sensitivity analysis. The proposed method provides a practical and generic way to build probabilistic decision support models by systematically exploiting expert knowledge. Suitable applications of this method include decision problems with multiple criteria, high uncertainty and limited data. We illustrate our method by applying it to a supplier selection case study in a large automobile manufacturer in Turkey.
ISSN:0957-4174
1873-6793
DOI:10.1016/j.eswa.2019.05.053