Tesis "Incorporating Preferences through an Evolutionary ε-Constraint Based Method"
Alumno: Giomara Lárraga Maldonado
Asesor: Ricardo Landa Becerra
Sinodales: Gregorio Toscano Pulido, Dr. Iván López Arévalo, Dr. Gerardo de la Fraga
Multiobjective optimization problems are common in the real world, for this reason, researchers of both Operational Research and Evolutionary Computation areas, have proposed many techniques for solving them. Preferences incorporation in multiobjective algorithms allows to obtain only the region of the Pareto Front preferred by the decision maker. This topic has been widely studied in the Operational Research area, however, further studies on this topic are required in the Evolutionary Computation area. Hybrid algorithms of Operational Research and Evolutionary Computation techniques have recently emerged as an alternative to improve deficiencies of both areas. In this thesis, a methodology to define the region of the Pareto front to be generated in an evolutionary multiobjective optimization algorithm is developed. Furthermore, this thesis also proposes a hybrid algorithm that, in addition to take advantage of this methodology to generate only a portion of the Pareto front, an operator based on ε-Constraint is used to improve its efficiency. The proposed algorithm was validated using metrics and functions taken from the specialized literature. Results indicate that the hybrid algorithm developed in this thesis is competitive with respect to representative algorithms of the state of the art.