An Interactive framework under conditions of uncertainty for multi-objective optimization

The last population generated by the evolutionary algorithm contains the best solutions; however, this population may be too extensive, thus difficulting the process of selecting the final solution that the decision-maker must perform. Therefore, a preference incorporation strategy must be integrated that approximates the interests of the decision maker to facilitate the solution's final choice. Different parameters are required to model the decision maker's interests, such as the objectives' weights to use the previously mentioned preference incorporation strategies. However, these values generally cannot be defined precisely by the decision maker, so ranges or intervals can be used to cover the uncertainty of these values. The decision maker's preferences can be considered before the execution of the evolutionary algorithm, at the end of the execution, or interactively during the algorithm's execution. This last method is the least studied because the process is more complex and slower than the a priori and a posteriori incorporation due to the intervention of the decision maker. Therefore, an interactive evolutionary framework has been proposed that uses preference disaggregation analysis and a chat-like interface. Then, through this proposal, the preferences of the decision maker can be efficiently incorporated, the number of tools that integrate this type of incorporation of preferences increases, and it demonstrates that the solutions converge before other types of articulation of preferences. Furthermore, with this proposal, the decision maker can see how the search moves in the solution space thanks to incorporating their preferences, thus facilitating the final choice of the solution.