Research On The Grey Wolf Optimizer And Its Application

With the increasing complexity of optimization problems in the real world,traditional optimization methods are difficult to solve.More and more researchers focus on using heuristic methods to find high-quality solutions to optimization problems.Evolutionary algorithms are meta-heuristics based on swarm intelligence.Because of their versatility and validity,they have become one of the mainstream methods for solving practical problems,which have been applied to the optimization problems in the fields of science,engineering and industry.They are global optimization methods with broad application prospects.The grey wolf optimizer is a swarm intelligence algorithm proposed by Seyedali Mirjalili in 2014 which mimics the social hierarchy and predatory behavior of grey wolves.It has the characteristics of fewer adjusting parameters and simple evolutionary formula.It has been applied to many practical optimizations.However,the algorithm also has the disadvantage of easily falling into local optimum.To tackle this problem,this paper proposes an improved grey wolf optimizer to enhance its convergence accuracy.Furthermore,the operators in the grey wolf optimizer are designed for the function optimization,which can not be directly used to solve the combinatorial optimization problem without discretization.Therefore,existing methods based on transfer function are studied to find the most suitable transfer function for the discretization of grey wolf optimizer.A binary grey wolf optimizer which can be applied to combinatorial optimization problem is proposed,and its optimization ability and robustness in the discrete space are tested by the benchmark instances of the multi-dimensional knapsack problem.Finally,the binary grey wolf optimizer is used to solve the feature selection,and a new approach for the problem is presented.The main research contents are as follows:1.By summarizing and comparing some existing improvement strategies for grey wolf optimzer,a scatter search strategy is proposed to improve the grey wolf optimizer,which utilizes the historical optimal information of individuals and enhances the information exchange between different individuals,based on the self-adjusting parameter,the convergence speed and accuracy of the alogrithm are improved.2.A binary grey wolf optimizer BGWO is proposed based on the comparison of five transfer functions for the discretization of evolutionary algorithms.In the process of individual evaluation of BGWO,a greedy strategy based repair and optimization method is introduced to handle the abnormal coding grey wolf individuals,and a new effective method for solving multi-dimensional knapsack problem is given.The performance of the algorithm is verified by the standard benchmark instances of the multi-dimensional knapsack problem.3.The binary grey wolf optimizer is applied to feature selection.A wrapper feature selection method based on binary grey wolf optimizer and KNN classifier is proposed.The classification effect is tested,analyzed and compared through UCI data set to validate the feasibility and effectiveness of the method proposed.