Due to its simple implementation,strong flexibility and high robustness,swarm intelligence optimization algorithms have attracted the attention of many researchers and have been widely used in various fields.In recent years,many novel swarm intelligence algorithms have been proposed,including the Grey Wolf Optimization(GWO)Algorithm,the Whale Optimization Algorithm(WOA)and the Grasshopper Optimization Algorithm(GOA).These algorithms are all implemented by simulating the predation and migration behavior of animals.Although these swarm intelligence algorithms have obvious advantages compared with classical swarm intelligence algorithms(such as genetic algorithm,particle swarm algorithm,etc.),they still have the problem of slow convergence and easy to fall into local optimum when optimizing practical problems.In this paper,the improved methods for each of the three swarm intelligence optimization algorithms are proposed,and these improved algorithms are successfully applied to the real optimization problems.The main research contents of this paper are as follows:(1)Based on the original GOA,an improved grasshopper optimization algorithm(IGOA)is proposed by introducing the opposition-based learning mechanism,Levy-flight mechanism and Gaussian mutation mechanism to improve the global and local search ability of the algorithm.In the IGOA,the Gaussian mutation mechanism is first used to increase population diversity and improve local search ability.Secondly,the Levyflight is used to improve the randomness of GOA and the ability to jump out of local optimum.Finally,the opposition-based learning is used to speed up the convergence of the algorithm.The results of the benchmark function optimization experiment show that IGOA has stronger global optimization ability than other swarm intelligence algorithms.The hybrid model IGOA-KELM based on IGOA also achieved ideal results in the experiment of predicting the company's financial stress.(2)Aiming at the problems existing in the WOA,this paper proposes an improved whale optimization algorithm(CCMWOA)based on chaotic initialization strategy,Gaussian mutation and chaotic local search strategy.In the CCMWOA,the chaotic initialization strategy is first used to generate high-quality initial populations,which is beneficial to improve the convergence speed of the algorithm.Second,the Gaussian mutation is used to ensure the population diversity of the algorithm in the iterative process.Finally,chaotic local search with a ‘shrinking' strategy is used to improve the local exploration ability of WOA.In the benchmark function optimization experiment,the global optimization performance of CCMWOA has been significantly improved compared with the original WOA.In the experiment of CCMWOA optimization three constrained engineering problems,the design scheme obtained by CCMWOA is obviously superior to other algorithms.(3)In order to better balance the global search ability and local search ability of the GWO,this paper proposes an improved grey wolf optimization algorithm(IGWO)based on the new hierarchical structure.In the new hierarchical structure,Beta will perform local search around the current optimal solution(Alpha),improve the local search ability of the algorithm,and speed up the convergence.The Omega will conduct a random global search of the entire solution space,avoiding the risk that the entire population will fall into local optimum.Experiments on the benchmark function show that the performance of IGWO is significantly better than the original GWO algorithm and other commonly used swarm intelligence algorithms.In addition,the IGWO was used to optimize the parameters of the KELM model,and a hybrid machine learning model IGWO-KELM was constructed and applied to the diagnosis of thyroid cancer.The experimental results show that the model can achieve 86.11% diagnostic accuracy,78.43% sensitivity and 92.43% specificity. |