Using Ordinal Regression for Interactive Evolutionary Multiple Objective Optimization with Multiple Decision Makers
[ 1 ] Instytut Informatyki, Wydział Informatyki, Politechnika Poznańska | [ P ] employee
2015
paper
english
- evolutionary multiple objective optimization
- interactive method
- group decision
- additive value function
- preference disaggregation
- NEMO
EN We present an interactive evolutionary multiple objective optimization (MOO) method incorporating preference information of several decision makers into the evolutionary search. It combines NSGA-II, a well-known evolutionary MOO method, with some interactive value-based approaches based on the principle of ordinal regression. We introduce several variants of the method distinguished by an elitist function indicating a comprehensive value that each solution represents to the group members. The experimental results confirm that all proposed approaches are able to focus the search on the group-preferred solutions, differing, however, with respect to both part of the Pareto front to which they converge as well as the convergence speed measured in terms of a change of utilitarian value of the returned solutions.
185 - 198
WoS (15)