A Multi-objective Optimization Evolutionary Algorithm using PCA and Gaussian Classifier method

The implementation of most current MOEAs directly adopt traditional genetic recombination operator such as crossover and mutation . In this study a new method based on the estimation of distribution algorithms for continuous multi-objective optimization problems with variable linkages, is proposed in which each generation, a promising area of search space, by a probability model, is built. On the promising area in search space, which is the same dominant solutions with better ranking; either clustering based on fuzzy techniques, or on the dominant solutions with best ranking, are performed; that in second method, a crowded tournament selection operator used to adjacent solutions, that solutions are too close together, be removed and the remaining points as the centers of clusters, are considered. Then, clustering based on nearest neighbors, is done. The principal component analysis algorithm, which neighbors, is done. The principal component analysis algorithm, which is best method to reduce data dimension linearly, has been used for modeling. New solutions have built from the model, based on a normal distribution is obtained. The Proposed method has been tested and the results of them are compared with NSGA-II method. The results show that this method is faster than previous methods and with fewer iterations and evaluation functions, better results are obtained.

کلیدواژه ها:

Gaussian Classifier Method ، Fuzzy c-means clustering ، Gustafson-Kessel clustering ، Multi-objective optimization ، PCA