Multi-Criteria Decision Making Using Bayesian Networks

In this thesis, we examine how to represent and combine the utilities elicited from multiple experts across multiple interdependent and uncertain criteria. We learn and formally represent those interdependencies using Bayesian network models and then incorporate the resultant probabilistic representations of utility as Bayesian decision networks. Treating utility dependencies explicitly has never been reported in conjunction to real-life multi-criteria decision-making problems. We present four different elicitation methods, namely, the Baseline method, the 4 Point method, AHP based method and causal discovery method to calculate the utility functions in the BDN and learn the dependencies among them.