Interactive Preference Incorporation in Evolutionary Multi-objective Engineering Design

Helping the human decision maker (DM) in finding the most preferred solution is the main purpose of Preference-based Multi-Objective Evolutionary Algorithm (P-MOEA). Interactive methods have been increasingly developed during the years because the preference information is less explicit than other methods. In addition, the DM can specify and correct his/her preference during the course of optimization. Our attention is to develop a new P-MOEA under the assumption that the DM must be available and willing to participate in solution process interactively. Our approach is hybridized between two well-known preference models, reference point and extended dominance approaches. Initially, pattern solution is approximated, Region of Interest (ROI) is identified by the means of user-supplied reference-point, and then a pruning mechanism called ASA approach is applied in order to filter out undesired solutions. The approach is evaluated using Component Redundancy Allocation Problem (RAP) in order to demonstrate the usefulness of our approach in practical context.