Research And Application Of Multi-objective Genetic Algorithm Based On Improved Banker Law And Clustering

Genetic algorithm was first proposed by Holland of the United States, then Goldberg summarized its basic frame. It is widely used in solving many problems of real life, such as production scheduling problem, combinatorial optimization, function optimization, image processing, robotics, machine learning, automatic control and so on. Genetic algorithm is good applied in dealing with single target problem, but many questions need to simultaneously satisfy multiple targets, so multi-objective genetic algorithm appeared.Multi-objective genetic algorithm based on Pareto is a research focus now, its main feature is that Pareto optimality can join into choice mechanism. Banker law is one of the effective methods to tectonic Pareto optimal solution sets. Clustering is a good method to maintain the diversity of the population. Combining it and banker law, we will get good optimizing operation results and operating efficiency.The main content includes the following aspects:1. The first chapter introduces genetic algorithm, multi-objective optimization problem, multi-objective genetic algorithm based on Pareto, research status of multi-objective genetic algorithm in domestic and foreign and so on.2. The second chapter introduces Pareto optimal solution sets, Pareto optimality boundary, the basic thought of banker law, and it detail described a example about finding non-dominated sets by banker law.3. The third chapter improves banker law by adding a vice banker in the original banker law. The vice banker participates comparison in each round, it can reduce the number of rounds. Comparing with the banker law which is not improved by experiments, we can find that the improved banker law is high efficiency. The clustering is introduced in the thesis, and it shows that the situation of optimal solution.4. The fourth chapter is that improved banker law and clustering is applied to the flow shop scheduling and water supply system. Algorithm keeps evolution group's diversity by clustering, and gets non-dominated sets by the improved banker law. Experiments show that the improved banker law and clustering can achieve better optimizing effect.