The Research And Application Of League Championship Algorithm

Swarm intelligence optimization algorithm is a research hotspot in recent years.lt has the advantages of simple mathematical theory, less dependent on background knowledge of optimization problems, no prior knowledge, wide applicability, good effect and so on. The advantages make it has wide application in data mining, image processing, and other areas of engineering practice. The league championship algorithm(LCA) is a swarm intelligece optimization algorithm based on iterative and has good calculation effect and calculation efficiency. This paper focuses on the research of the LCA, including improvements the drawback of the algorithm and extension the algorithm into multi-objective league championship algorithm that can solve the multi objective optimization problem. At last, this paper verify the validity of improved method and extensional algorithm by a large number of experiment. The research work mainly includes the following four points:The first, by summarizing the classical swarm intelligence optimization algorithm and analysis the LCA model discovery the LCA exist some drawbacks. And the drawbacks include that the LCA calculation effect and efficiency is strongly dependent on the parameter value setting, and algorithm missing an important strategy to maintain population diversity and leading to easy to fall into local optimal solution.The second, first, detection algorithm running state strategy is proposed in this paper according to the characteristics of LCA and the strategy is added to the LCA. Second, adaptive parameter adjustment method is proposed in this paper in order to reduce the impact of the parameter value setting on the LCA running effect and efficiency. The method obtain the specific influence of the parameter value setting on the algorithm through the theoretical analysis and experimental verification, and then start the adaptive parameter adjustment according to the result of the detection algorithm running state, and avoid the users trouble of set the parameters when using LCA. At last, knockout system proposed in this paper in order to reduce the probability of the LCA falling into local optimal solution, and the system idea is comes from the realistic sport league championship. So that enhancing the global search ability of the LCA. The experiment show that the improved algorithm proposed by this paper improves the global optimization ability and simplifies the use of the LCA.The third, this paper summarizes the general framework and several important strategy of the multi-objective evolutionary algorithm by analysis the model of classical multi-objective evolutionary algorithm. And according to the general framework and combining with the characteristics of LCA, this paper presents the basic flow of the multi-objective league championship algorithm(MOLCA).The fourth, the LCA is a single objective optimization algorithm and we must modify the part steps of the LCA to meet the characteristics of multi-objective algorithm. If LCA direct application to the general framework of multi-objective evolutionary algorithm will make the algorithm to lack of maintainance population diversity strategy, and lead to the algorithm easy to fall into local optimal solution set. To slove the situation, optimal solution diffusion strategy is proposed in the paper. The strategy diffuses optimal solution for several solution in order to make the decision space for a more comprehensive search. At last, the automatic termination of the MOLCA is proposed in the paper. The experiment show that the MOLCA have better pareto optimal solution set than the current relatively excellent multi-objective evolutionary algorithm.