Improved Research And Application Of Seagull Optimization Algorithm

With the progress and development of artificial intelligence technology,the optimization problem has not only been focused on in the field of mathematics,but also received widespread attention in the social fields of industrial manufacturing,driverless,intelligent computing and so on.The essence of the problem can be described as the organic combination of the decision variables of the problem under limited conditions,and then the feasible solution or the best solution of the problem can be obtained.In view of the fact that traditional mathematical methods cannot effectively solve the high-dimensional multi-constraint optimization problems in real life,scholars have shifted their research focus to the meta-heuristic algorithm based on swarm intelligence,which makes use of the characteristics of flexible structure,high stability,and strong optimization ability under complex constraints to ingeniously make up for the performance shortcomings of traditional mathematical methods in optimization problems.Seagull optimization algorithm is a meta-heuristic algorithm based on swarm intelligence,which has emerged in recent years.This algorithm obtains the global optimal solution of the optimization problem through bionic experiments on the migration and attack behavior of natural seagulls.Seagull algorithm has simple principle,flexible structure,high efficiency and easy implementation.It has great application potential and wide application scenarios in resource scheduling,spatial location,fault detection,feature selection and other fields.In order to better solve large-scale and multi-constraint complex optimization problems and further explore the application scenarios of swarm intelligence optimization algorithms,this thesis takes Seagull algorithm(Seagull Optimization Algorithm)as the research object,proposes two improved variants for different types of target optimization problems,and has achieved the following research results:(1)A seagull algorithm(L-SOA)based on good point set mapping and dual hybrid strategy is proposed for single objective engineering optimization.The initial position of the population is disturbed by using the good point set mapping,which improves the quality of the initial solution of the population;The linear weight in the original algorithm is replaced to enhance the ability of the population to get rid of local adverse solutions;The Levy flight and adaptive walk strategy are integrated into the original population update formula,maintaining the balance between the global exploration and local development capabilities of the algorithm.On the selected eight performance test functions,the optimization differences between L-SOA and other algorithms are compared from multiple perspectives.Finally,L-SOA algorithm is used to solve two classical engineering optimization problems,and the best solution results of all the comparison algorithms are obtained.(2)A binary seagull algorithm(BSOA)based on sigmoid factor and population reduction strategy is proposed for multi-objective feature selection.The sigmoid function is used to convert the position vector in the single-objective seagull optimization algorithm into binary discrete values as the input of the feature selection problem,which improves the efficiency of the algorithm.Secondly,the population reduction strategy is introduced into the iterative optimization process to save the time cost of the algorithm.Finally,a performance comparison test was conducted on 10 UCI classic data sets,and BSOA obtained higher classification accuracy and fewer selected features than other binary variants.