Title: Extracting decision rules using version spaces in a qualitative approach to multi-attribute decision aid

Speaker: Miguel Couceiro, (Orpailleur Team, LORIA, Inria Nancy - Grand Est, France)

Short bio: Miguel Couceiro received his degree of Doctor of Philosophy from the University of Tampere, Finland, in 2006, and his Habilitation degree in Computer Science from the University Paris-Dauphine, France, in 2013. He was a postdoctoral fellow at the University of Luxembourg (2007-2012) and an Associate Professor at University Paris-Dauphine (2012-2014). Currently he is a full Professor of Orpailleur Team at LORIA (CNRS – Inria Nancy Grand Est – University of Lorraine).
His research interests can be found in discrete mathematics, theoretical computer science, multicriteria decision aid and artificial intelligence. His earlier work focused on function theory, including aggregation theory, clone theory, multiple-valued logic. His recent works are pertaining to multicriteria decision making, in particular, preference modelling, reasoning and learning (with particular emphasis on aggregation, decomposition and reconstruction techniques). He has more than 100 peer-reviewed papers in international journals and conference proceedings, and he has co-organized several international conferences and colloquia.

Abstract: We consider a lattice-based model in multi-attribute decision making, where preferences are represented by global utility functions that evaluate alternatives in a lattice structure. Essentially, this evaluation is obtained by first encoding each of the attributes (nominal, qualitative, numeric, etc.) of each alternative into a distributive lattice, and then aggregating such values by lattice functions. As shown by Grecco et. al, and independently by Bouyssou et. al, such a model is equivalent to the rule based decision model.

We formulate version spaces within this model as solutions of an interpolation problem and present their complete descriptions accordingly. As it turns out, up to 3 attributes this interpolation problem is solvable in polynomial time if there are at most 3 attributes, otherwise it is NP-complete. If time allows, we will illustrate these results with a concrete example, namely, a recommender system for employees based on their psychological records throughout a year, and present the rules that can be extracted from each of the solutions presented.