The NN-DM method: an artificial neural network model for decision-makers preferences

Abstract

This work presents a methodology based on the multi-attribute utility theory to approximate the decision-makers utility function: the Neural Network Decision-Maker method (NN-DM method). The preference information extracted from the Decision-Maker (DM) involves ordinal description only and is structured by a partial ranking procedure. An artificial neural network is then constructed to approximate the partial ranking reproducing the DMs preferences in a specific domain. The NN-DM method is suitable in situations in which a recurrent decision process must be performed considering different sets of alternatives and the same DM. A hybridization between the NN-DM method and the Interactive Territory Defining Evolutionary Algorithm (iTDEA) is also developed in this work. Considering the same amount of preference information, the NN-DM method is able to construct a model for the DMs preferences to guide iTDEA. Henceforth, no further queries are required from the DM related to similar decision-making problems. Additionally, an Interactive Non-dominated Sorting algorithm with PreferenceModel (INSPM) based on NSGA-II is proposed. A slight modificationin the diversity maintenance strategy inside NSGA-II enables INSPM to distinguish preferable regions within an estimate of the Pareto-optimal front. A parameter allows the control of the preferable regions density and provides from fronts in which there is no interference from the DM until fronts in which the preferred solution is apparent. In all situations the Pareto-front extent is guaranteed. Finally, the NN-DM method is adapted to find a model for the DMs preferences in a polymer extrusion process. The DMs requirement is filling a matrix expressing the preferences considering ordinal comparisons. The NNDM method is able to provide a model which sorts the alternatives from the best to the worst one according to the DMs preferences in a real scenario.