Weighting models using cross-sorting methods for collaborative decision support in projects.

Summary In project management, weighting procedures are associated with the evaluation of parameters involved in the decision-making process. In spite of this issue, field tools are very often not adapted to project requirements. Some even lead to inaccurate results. At the same time, the solutions available in the research world are not always flexible enough to be used immediately in the field. This thesis provides some answers both on a theoretical and practical level. It is part of the more specific framework of so-called cross-sorting tools. This designation includes any method reducing the problem to a set of binary comparisons of the elements to be weighted in order to simplify the task of the decision group. Our contributions consist first of all in a greater flexibility in taking into account the opinions of the decision makers. A general mathematical formulation of the problem is proposed as well as perspectives, based on fuzzy logic, for a better management of the imprecision inherent in the judgment. Our contribution also concerns the detection and treatment of the inconsistency induced by the aggregation of binary comparisons which often contain redundant information without necessarily being concordant. An original coherence indicator is proposed as well as an iterative procedure to improve this coherence. Moreover, the introduction of the notion of voting entity allows to consider a larger variety of voting strategies and to target the assistance brought to the decision group. Finally, on a practical level, an original concept of asynchronous and distributed cross sorting platform is proposed. It offers great flexibility in the design and implementation of a cross sorting procedure in a project. A computer model has been developed in order to validate a part of its functionalities.