A Variable Neighborhood Immune Algorithm Based On Forecasting Choice For Dynamic Multi-objective Optimization

In the field of intelligent algorithm,research in solving optimization problems with evolutionary algorithm has gained fruitful results and that with swarm intelligence algorithm,such as the particle swarm algorithm and ant colony algorithm,has also made some achievements.Artificial immune algorithm,which is based on natural immune system,is increasingly showing its advantages in solving optimization problems with the deepening of the theoretical research.This paper aims to solve dynamic multi-objective optimization problems by adopting a new variable neighborhood immune optimization algorithm based on forecasting choice.Dynamic multi-objective optimization problem,as a basic problem,is commonly seen in scientific research and engineering practice and it has a promising future in terms of its extensive application and research significance.In solving dynamic multi-objective optimization problems,we need to take into account that the solutions of different objective functions would conflict with each other and the objective functions and constraint functions,namely the optimal solution,may vary with time.Currently,most dynamic multi-objective optimization algorithms,generally based on some successful static multi-objective optimization algorithms,tend to adopt an auxiliary strategy such as random initialization population strategy to restart the algorithm and adapt itself to the environmental changes.Although this method can get the optimal solution set,it meanwhile increases the randomness as well as reduces the rate of convergence of the algorithm.Some algorithms adopt prediction strategies to forecast the optimal solution of the new environment when an environmental change occurs,but there are some limitations due to the single prediction algorithm they use.This paper was aimed at solving dynamic multi-objective optimization problems with a variable neighborhood immune algorithm based on forecasting choice and it mainly includes the following three parts.Firstly,gives a detailed introduction of dynamic multi-objective optimization problems,including its mathematical models,the related concepts and definitions,as well as a status report on research in this field at home and abroad.According to the research,the author designed a basic framework and process for addressing dynamic multi-objective optimization problems.Secondly,presents a novel variable neighborhood immune optimization algorithm based on forecasting choice to solve dynamic multi-objective optimization problems.This algorithm mainly includes three prediction models,and introduces a heuristic algorithm to help select one from the three models.Whenever the environment changes,the algorithm has a relatively high probability to select the prediction model which achieves optimal prediction effect and best represents the trend of environmental changes to initialize the population.The author adopted immune optimization algorithm,which combines immune network sorting with variable neighborhood strategies,as the main algorithm to work out the optimal solution in the relative static environment between two neighboring changes.Finally,introduces two benchmark test functions and algorithm performance metrics involved in dynamic multi-objective optimization problems.After running the algorithm this paper proposes on these test functions,the author made a comparative analysis of the experiment results with the results obtained from other algorithms,to verify the validity,distribution and uniformity of the Pareto solution set we obtained with the algorithm proposed in this paper.In conclusion,this paper has three innovation points.Firstly,grey prediction theory was introduced into solving dynamic multi-objective problems.Secondly,the paper proposed a novel prediction algorithm,which drew ideas from hyper-heuristic algorithm.Finally,to obtain the optimal solution,the author integrates variable neighborhood immune optimization algorithm with prediction algorithm.Through experiments and the results,we can know the variable neighborhood immune algorithm based on forecasting choice for dynamic multi-objective optimization has achieved good results.