| In an organization, executive decisions often are achieved through a process that involves groups of people instead of isolated individuals and, most of the time, the problems involved are highly complex. The complexities of the problems derive from having multiple criteria and multiple decision makers. In these environments, executive decision makers require appropriate systems for supporting their beliefs. Such a system should integrate a variety of communication mechanisms for which group members can follow their individual objectives and resolve conflicts. It should also take into account that reasoning is justifiable; future information may cause another alternative to be more preferable than the solutions at the moment.; In general, group decision making techniques with multiple attributes are based on the Analytic Hierarchy Process (AHP), utility theory, ordinal and cardinal approaches. For practical reasons, the cardinal approach which involves scoring candidates is the most appropriate technique to be applied to organizational risk profiling. However, the problem with the cardinal approach is that it orders the alternatives according to the distances from either the positive-ideal solutions or the negative-ideal solutions, which are determined by the respective attributes. By doing so, the solutions to the problem are prejudiced by the fact that the alternatives that receive very high scores from one expert could be the most favorable alternatives, regardless of the opinions from the rest in the group. A new clustering method is needed that overcomes these deficiencies and that is practical to implement.; This dissertation presents a new decision making algorithm called ClusterGroup, which is developed for dealing with multi-criteria group decision making. The main concept of this new approach is to cluster the similar opinions of executives or experts into groups. Then, the cluster with the majority of the opinions will outweigh over the rest and is considered as representative of the group in the final calculation. This new technique will be a computer-assisted multi-criteria group decision making process, for which the identity of each expert is concealed from their decisions. The purpose of doing so is to avoid socio-psychological pressures and to prevent the dominance of one member over the others.; The dissertation offers the development of the ClusterGroup algorithm and illustrates the method with both real world and synthetic data sets. Comparisons with traditional approaches are discussed. Potential applications to the aviation risk management domain are addressed. |
