Decision Making Based on Past Problem Cases

This paper deals with the generation of an evaluation model to be used for decision making. The paper proposes the automated selection of past problem cases and the automated synthesis of a new evaluation model, based on the cumulative experience stored in a knowledge base. In order to select the most promising past evaluation cases we propose the use of two metrics: their proximity to the new case and the degree of success. To add flexibility, we allow the user to express his preference on these two factors. After having selected a group of the most promising past evaluation cases, a method for deriving a new evaluation model, i.e. the weights and the scales of the attributes, is presented. The method covers both numerical and nominal attributes. The derived model can be used as a starting point for an interactive evaluation session. The overall process is illustrated through a real world situation, concerning the choice of 1-out-of n candidate ERP products for an enterprise information system.