Learning And Predicting Individual Preferences In Multicriteria Decision Making With Neural Networks Vs. Utility Functions

Abstract

This paper reports an empirical investigation into the performance of neural network technique vs. traditional utility theory-based method in capturing and predicting individual preference in multi-criteria decision making. As a universal function approximator, a neural network can assess individual utility function without imposing strong assumptions on functional form and behavior of the underlying data.  Results of this study show that in all cases, the predictive ability of neural network technique was comparable to the multi-attribute utility theory-based models.