Learning of aggregation models in multi criteria decision making

Abstract

Generalized attitudinal Choquet integral (GACI) is a recent aggregation operator that subsumes a multi- tude of aggregation operators, including both linear as well as non-linear and exponential integrals. In this study, against the background of preference learning, we use the GACI operator to represent the util- ity function of a decision-maker (DM), and learn its parameters. The exemplary preference information in the form of pair-wise comparisons of alternatives constitutes the training information. More specifically, given the exemplary pairwise choices of a DM, we present an approach to infer the unique preference model of the DM, in terms of the parameter values of GACI operator. We test our approach on standard datasets, and the prediction performance is compared with state-of-the-art methods.