Heuristic Salp Swarm Algorithm And Application Research

The salp swarm algorithm(SSA)is a novel metaheuristic optimization algorithm that simulates the foraging behaviour of a community of salp swarm in the deep sea.The algorithm has a simple structure,is easy to understand and has a strong search capability,and is widely used in engineering,economics,medicine and other fields.As the SSA was studied in depth,researchers found that the algorithm suffers from slow convergence,low search efficiency and a tendency to fall into premature maturity at a later stage.This thesis analyzes and improves the shortcomings of the Salp Swarm Algorithm,and applies the improved algorithm to some complex optimization problems,with the aim of further improving the theory of SSA algorithm,optimizing its performance and widening its application scope.The main work of this thesis is as follows:(1)A neighbourhood centre of gravity backward learning based salp swarm algorithm(NSSA)is proposed to enhance the global search capability of the SSA,better balance its global and local search capability and improve its convergence accuracy.It was used to train an extreme learning machine,tested on classification and regression datasets,and compared with the results of other algorithms,and NSSA was found to have better performance.It also performs well in predicting dew point temperatures.(2)A complex-valued encoding salp swarm algorithm is proposed,which adopts the idea of complex-valued encoding double chaperone set to encode the individuals of the salp swarm algorithm,increasing the diversity of the population,enhancing the global exploration ability of the algorithm,avoiding the algorithm from falling into local optimum prematurely,and overcoming the shortcoming of insufficient accuracy in the later stage of the optimization search.The experimental results demonstrate the superior performance of the proposed complex-valued encoding of the salp swarm algorithm.(3)The complex-valued encoding of the salp swarm algorithm was used to train a feed-forward neural network to find an optimal set of values for the weights and deviations to achieve minimum error.The complex-valued salp swarm algorithm has better performance when compared with other algorithms on the test dataset.