A Methodology for Modeling Influence Diagrams: A Case-Based Reasoning Approach

Abstract

In this paper, a case-based reasoning approach to build an influence diagram is described. Building an influence diagram in decision analysis is known to be a most complicated and burdensome process. To overcome such a difficulty, decision class analysis is suggested, which treats a set of decisions having some degree of similarity as a single unit. This research suggests a case-based reasoning approach as a methodology to analyze a class of decisions. The candidate influence diagrams are retrieved from a set of similar influence diagrams, a case base. They are combined and modified by the node classi-fication tree and DMs preference for the given decision problem. For such a purpose, the case representation and retrieval process is explained with the adaptation process. We suggest using two measure, the fitness and garbage ratio for the case retrieval process. The basic concept of decision class analysis and case-based reasoning is very similar so the case-based reasoning approach is believed to be a better methodology to implement a decision class analysis.