Pareto Dominance-based MOEA with Multiple Ranking methods for Many-objective Optimization
Vikas Palakonda, Samira Ghorbanpour, Rammohan Mallipeddi · 2018
Pareto Dominance-based Multi-objective Evolutionary algorithms (PDMOEAs) have issues while handling many-objective optimization problems (MaOPs) due to the lack of selection pressure provided by the Pareto dominance to guide the search process towards the convergence. Hence, most of the PDMOEAs proposed rely on additional selection criterion to establish preferences between the solutions. In this paper, we propose a PDMOEA with multiple ranking methods (PDMOEAMR), an extension to the proposed PDMOEA with ranking methods for MaOPs which assigns priority rank based upon Ranking methods and niche radius. In PDMOEA with ranking methods for MaOPs, ranking methods such as Average rank (AR) in PDMOEA-AR, and weighted sum of objectives (WS) in PDMOEA-WS, are used. Instead of using two ranking methods separately, in the proposed PDMOEA_MR, both the ranking methods AR and WS are incorporated into a common framework and a different strategy is adopted to assign priority rank. The performance of proposed method is analyzed on 16 test problems.