loading page

A Dynamic Multi-Objective Evolutionary Algorithm based on Classification Prediction of Decision Variables
  • Shenghui Zhang,
  • Er-chao Li
Shenghui Zhang
Lanzhou University of Technology

Corresponding Author:[email protected]

Author Profile
Er-chao Li
Lanzhou University of Technology
Author Profile

Abstract

Evolutionary algorithms must be able to quickly and effectively approach the optimal frontier, locate the best collection of solutions, and sensitively perceive and react to environmental changes in order to successfully solve dynamic multi-objective optimization problems. Most dynamic multi-objective optimization algorithms currently in use evolve the decision variables uniformly, without taking into account how environmental changes may affect the decision space, without effectively addressing decision variables with different characteristics, and without being able to handle the useful information produced during the evolutionary process. In order to address the problem, this paper proposes a dynamic multi-objective evolutionary algorithm based on the classification prediction of decision variables (DMOEA-DVCP). Firstly, the degree of influence of environmental changes on each dimension of the decision space is measured by calculating the degree of change of each dimension of the decision variables; secondly, the decision variables are classified into three categories according to the strength of change, i.e., strong change, weak change and change between strong and weak strength; finally, three distinct response strategies are employed based on classification, the decision variables of each dimension are initialized adaptively, and the newly generated individuals are considered the initial population in the new environment. To verify the effectiveness of DMOEA-DVCP, it is compared with four representative algorithms on 10 typical dynamic test functions, and the experimental results show that the DMOEA-DVCP algorithm has obvious advantages in solving dynamic multi-objective optimization problems.