A Novel Integration of MCDM Methods and Bayesian Networks: The case of Incomplete Expert Knowledge

In this study, we propose an effective integration of multi criteria decision making methods and Bayesian networks (BN) that incorporates expert knowledge. The novelty of this approach is that it provides decision support in case the experts have partial knowledge. We use decision-making trial and evaluation laboratory (DEMATEL) to elicit the causal graph of the BN based on the causal knowledge of the experts. BN provides the evaluation of alternatives based on the decision criteria which make up the initial decision matrix of the technique for order of preference by similarity to the ideal solution (TOPSIS). We then parameterize BN using Ranked Nodes which allows the experts to submit their knowledge with linguistic expressions. We propose the analytical hierarchy process to determine the weights of the decision criteria and TOPSIS to rank the alternatives. A supplier selection case study is conducted to illustrate the effectiveness of the proposed approach. Two evaluation measures, namely, the number of mismatches and the distance due to the mismatch are developed to assess the performance of the proposed approach. A scenario analysis with 5% to 20% of missing values with an increment of 5% is conducted to demonstrate that our approach remains robust as the level of missing values increases.