Laboratoire de Génie Informatique et d’Automatique de l’Artois

Multi-criteria decision making is the study of identifying and choosing alternatives to find the best solution based on different criteria and considering the decision makers’ expectations. However, the expert assessments are sometimes expressed imperfectly. Belief function theory can then provide more flexible and reliable tools to manage different types of imperfection. Thus, in this thesis, we are interested in multi-criteria decision making in an uncertain framework by extending the Analytic Hierarchy Process (AHP) method to the belief function framework. After presenting the theoretical foundations of the AHP method, we proposed an approach that reduces the number of pair-wise comparisons by judging subsets of criteria and alternatives. In addition, we examined the dependence between the criteria and alternatives. In this case, the uncertainty is given in terms of conditional mass distributions. Another part of the work provides critical concerning the pair-wise comparison process. For this purpose, we proposed two approaches. The first expert judgment elicitation method is based on mass distributions, while the second one is based on preference relations. In this context, we have introduced a model that is able to generate quantitative mass distributions from preference relations. Thus, we have developed a multi-criteria decision making method that imitates human reasoning. This method gives better and more robust results than existing approaches.