User profiling based on multi-criteria analysis: the role of utility functions

It is shown in this work that the incorporation of a Multi-criteria methodology prior to the application of a clustering algorithm constitutes a fundamental step for the formation of more accurate user profiles in terms of compactness and separation of the groups.

Abstract

Personalized profiles that describe user behaviour and preferences are encountered in many applications, ranging from recommender systems or user agents (Web browsers) to one-to-one marketing. User profiling is a crucial step in providing accurate product recommendations to the end users. Once groups of people with similar preferences are identified, information, products, or services tailored to group's needs, can be delivered. The aim of this paper is to clarify the role of utility functions in the formation of user profiles. Thus, the clustering behaviour of customers' preferences is investigated by means of clustering algorithms and preference modelling. It is shown in this work that the incorporation of a Multi-criteria methodology prior to the application of a clustering algorithm constitutes a fundamental step for the formation of more accurate user profiles in terms of compactness and separation of the groups.