Preference Incorporation in Evolutionary Multiobjective Optimization

After using Evolutionary Algorithms (EAs) for solving multiobjective optimization problems for more than two decades, the incorporation of the decision maker's (DM’s) preferences within the evolutionary process has finally become an active research area. In fact, EAs have demonstrated their effectiveness and efficiency in providing a well-converged and well-distributed approximation of the Pareto front. However, in reality, the DM is not interested in discovering the whole Pareto front rather than approximating the portion of the front that best matches his/her preferences, i.e., the Region Of Interest. For this reason, many new preference-based Multiobjective Optimization EAs (MOEAs), which are mostly variations of existing methods, have been recently published in the specialized literature. The purpose of this chapter is to summarize and organize the information on these current approaches in an attempt to motivate researchers to further focus on hybridizing between decision making and evolutionary multiobjective optimization research fields; consequently facilitating the DM's task when selecting the final alternative to realize. Hence, a summary of the main preference-based MOEAs is provided together with a brief criticism that includes their pros and cons. Furthermore, we propose a classification of such type of algorithms based on the DM's preference information structure. Finally, the future trends in this research area and some possible paths for future research are outlined.