A novel hybrid evolutionary algorithm for solving multi-objective optimization problems

Abstract

This paper proposed a novel hybrid evolutionary algorithm for solving the multi-objective optimization problems (MOPs). The algorithm uses the idea of simulated annealing to combine co-evolution with genetic evolution, set several model sets according to the principle of "model student", and employs ε-dominant and crowding distance sorting to search the excellent population. In the co-evolution model, cluster analysis is used to classify the model set, and the estimation of distribution algorithm (EDA) is used to establish the probabilistic model for each class. The individuals are generated by sampling through the probabilistic model; in genetic evolution model, populations evolve based on the model set. This algorithm takes full advantage of the global and local search abilities, and makes comparison with the classical algorithm NSGA-II, the experimental results show that our algorithm for solving the multi-objective problem has better convergence and distribution. © 2012 Springer-Verlag.