An improved bagging ensemble surrogate-assisted evolutionary algorithm for expensive many-objective optimization

摘要

When the surrogate-assisted evolutionary algorithm is used to solve expensive many-objective optimization problems, the surrogate is used to approximate the expensive fitness functions. However, with the increase of the number of objectives, the approximate error of the surrogate will accumulate gradually, and the computational cost will also increase sharply. This paper proposes an improved bagging ensemble surrogate-assisted evolutionary algorithm (IBE-CSEA) to solve these problems. An ensemble classifier is used to classify the offspring instead of building the surrogate to approximate the fitness function of each objective. Firstly, a group of classification boundary individuals are selected one by one from the individuals evaluated by the expensive fitness function. All the individuals evaluated by the expensive fitness function are divided into two categories; Secondly, the individuals in these two categories are divided into the training set and the test set. The training set is used to train an improved bagging ensemble classifier. The test set is used to calculate the reliability of the classification; Finally, the classification results and the reliability are used to select the promising individuals for expensive fitness function evaluation. Compared with the current popular surrogate-assisted evolutionary algorithm, IBE-CSEA algorithm is more competitive.