Dynamic multi-objective optimization algorithm based on weighted differential prediction model

In this paper, a new algorithm for solving dynamic multi-objective optimization problems(DMOPs) is proposed. Most of the traditional dynamic multi-objective optimization algorithms will make predictions based on the overall average evolutionary direction of the population, which is hardly applicable to problems where the solution set and frontier do not vary with the environmental rules. In this paper, a dynamic multi-objective optimization algorithm based on weight difference prediction model is designed to solve such problems. The algorithm contains a weighted differential prediction strategy, and a differential model is built for each individual using the weights to predict the initial population after environmental changes. With this approach, each individual in the population can be made to respond quickly to environmental changes. We used three classical comparison algorithms to conduct experiments on a series of test problems. The experimental results show that the WD-MOEA/D algorithm can significantly improve the dynamic optimization performance and is effective in solving different types of dynamic problems.