15h30 - 15h55
Preference learning in multicriteria decision analyses: UTASTAR-T, a novel time-series-based optimization method
Multicriteria decision analysis (MCDA) methods are used in decisions involving a set of alternatives according to multiple criteria. Preference learning in MCDA infers the preferences of a decision-maker provided indirectly by him/her.They generally use a linear programming formulation to determine piecewise linear marginal value functions.
In these methods, each criterion is associated with a single value, such as the criterion performance average in a given period of time or the criterion's last value available. Although this is appropriate in many situations, decisions are often made considering some characteristics of the criteria's time-series (summary measures such as the average and tendency). For instance, in medium- or long-term investment decisions, the tendency of the criteria is as relevant as their average.
Thus, we propose UTASTAR-T, an extension of the well known MCDA method, UTASTAR, to learn the DM's preferences in a context where the characteristics of the time-series criteria are considered simultaneously. The proposed method was tested and validated using actual data for the assessment of ten countries, using three criteria, life expectancy at birth, education, and gross national income per capita, while taking into account the criteria’s average and tendency.
15h55 - 16h20
MCDA decision process for strategic environmental and social assessment at the regional planning level
An eight steps decision process are discussed. They provide interested parties with a collaborative and contributive framework to understand and organize the issues raised by the SESA to be addressed. They also provide the accountable public authorities the information needed to clarify their understanding of the problem and to make the decision.