Abstract Revision of Qualitative Multi-Attribute Decision Models.pdf

881 Revision of Qualitative Multi-Attribute Decision Models Martin ?nidar?i ? * Marko Bohanec** Jo?ef Stefan Institute Department of Intelligent Systems Ljubljana, Slovenia *Email: martin.******@ **Email: marko.******@ Abstract Data-driven revision of decision models is defined as follows: given an existing model and a set of data items, revise the model to match the data items. In this paper, we propose and experimentally evaluate a method for the revision of qualitative hi erarchical multi-attribute models in DEX methodology . The revision method is automatic, but limited to the mod ification of utility func tions. Using a simple experimental model and simulated data, we show that the met hod is valid and that it improves the cl assification accuracy of the models. Keywords Decision support, qualita tive multi-attribute decision mode ls, data-driven model revision 1. INTRODUCTION Decision models are an essential part of decision analysis. Usually a lot of effort is dedicated to the construction of a suitable and useful model. Expert knowledge and data that describes the decision problem or known solutions are bined into a model. The use of the model depends on the characteristics of decision problem. Some models are used only once, when a difficult decision has to be thoroughly analysed (BRS 1997, Greenberg et al. 2002). On the other h