Integrating Preference by Means of Desirability Function with Evolutionary Multi-objective Optimization
Zhenhua Li, Hai‐Lin Liu · Intelligent Automation & Soft Computing · 2014
Desirability function is a mathematically simple description of decision maker's preference. A desirability function transforms objective function to a scale-free desirability value, which actually measures the decision maker's satisfaction with the objective value. In this paper, we utilize desirability functions to express decision maker's preference to specific regions with an objective. These desirability functions are integrated into evolutionary multi-objective algorithms to generate a uniformly distributed set of Pareto solutions in desirability space. The corresponding images in objective space of this set of solutions are exactly the decision maker's preferred solutions. The experimental results show the effectiveness of this approach.