Interactive evolutionary multiple objective optimization algorithm using a fast calculation of holistic acceptabilities

We propose a novel algorithm, called iMOEA-HA, for interactive evolutionary multiple objective optimization. It combines, in an original way, the state-of-the-art paradigms postulated in evolutionary multiple objective optimization and multiple criteria decision aiding. When it comes to the incorporated evolutionary framework, iMOEA-HA implements a steady-state, quasi decomposition-based model with a restricted mating pool. To drive the population towards the Decision Maker's (DM's) most preferred region in the Pareto front, iMOEA-HA exploits the DM's pairwise comparisons of solutions from the current population and assesses all solutions as imposed by their holistic acceptability indices. These acceptabilities are built on the probabilities of attaining a certain number of the most preferred ranks. We evaluate the proposed algorithm's performance on a vast set of optimization problems with different properties and numbers of objectives. Moreover, we compare iMOEA-HA with existing interactive evolutionary algorithms that can be deemed its predecessors, proving its high effectiveness and competitiveness.