Abstract

In this work we introduce a novel approach for bringing collective intelligence methods into the optimization process carried out by evolutionary multi-objective optimization algorithms. Expressing preferences from a unique or small group of decision makers may raise unilateral choices issues and poor hints in terms of search parameter. The extension of the non-dominated sorting genetic algorithm II (NSGA-II) and S-metric selection algorithm (SMS-EMOA) to include collective preferences works on refining users' preferences throughout the optimization process to improve the reference point or fitness function. Supported by dynamic group preferences, the interactive algorithms - which we called CI-NSGA-II and CI-SMS-EMOA - aggregate consistent collective reference points to enhance multi-objective results and highlight the regions of Pareto frontier that are more relevant to the decision makers. The algorithms performance are tested on scalable multi-objective test problems and a real-world case of resource placement.