MOSA/D-O and MOSAD/D-O-II: Performance analysis of decomposition-based algorithms in many objective problems
Manuel Vargas-Martínez,
Nelson Rangel-Valdez,
Eduardo Fernández,
Claudia Gómez-Santillán,
Gilberto Rivera,
Fausto Balderas
Affiliations
Manuel Vargas-Martínez
División de Estudios de Posgrado e Investigación Tecnológico Nacional de México, Instituto Tecnológico de Ciudad Madero, 89440, Ciudad Madero, Tamaulipas, Mexico; Corresponding author.
Nelson Rangel-Valdez
CONACyT Research Fellow at Graduate Program Division, Tecnológico Nacional de México, Instituto Tecnológico de Ciudad Madero, 89440, Ciudad Madero, Tamaulipas, Mexico
Eduardo Fernández
Facultad de Contaduría y Administración, Universidad Autónoma de Coahuila, 27000, Torreón, Coahuila, Mexico
Claudia Gómez-Santillán
División de Estudios de Posgrado e Investigación Tecnológico Nacional de México, Instituto Tecnológico de Ciudad Madero, 89440, Ciudad Madero, Tamaulipas, Mexico
Gilberto Rivera
División Multidisciplinaria de Ciudad Universitaria, Universidad Autónoma de Ciudad Juárez, 32579, Cd. Juárez, Chihuahua, Mexico
Fausto Balderas
División de Estudios de Posgrado e Investigación Tecnológico Nacional de México, Instituto Tecnológico de Ciudad Madero, 89440, Ciudad Madero, Tamaulipas, Mexico
In recent years, many-objective optimization problems (MaOPs) have been challenging. Classically, algorithms obtain first the Pareto front (PF). Next, a decision maker (DM) can choose the best solutions according to their preferences in the region of interest (ROI). However, the DM effort increases with objectives in MaOPs. For this reason, this paper proposes a new C++ software based on simulated annealing, decomposition, differential evolution, and outranking relations. MOSA/D-O and MOSA/D-O-II algorithms from the software can pressure toward the ROI at run-time in MaOPs. Both algorithms were tested with DTLZ and WFG benchmarks, showing promised performance in an ad hoc indicator for 5 and 10 objectives with 10 DMs.
