Improved Jellyfish Search Optimizer For Multi-Application Scenarios

Optimization problems widely exist in the fields of scientific research and technical engineering,however,some of these problems are difficult to solve due to the characteristics of high dimensionality and multiple constraints,and swarm intelligence algorithms can be utilized to effectively solve such optimization problems.Swarm intelligence algorithms complete mathematical modeling by observing and simulating the behavior of various organisms in nature.The modeling process mainly includes randomly generating populations,completing the search for solutions through updates and iterations,and finally achieving the solution goal.Jellyfish search optimizer is a novel swarm intelligence algorithm inspired by the foraging behavior of jellyfish,which has the advantages of strong applicability,simple structure and strong global search ability.Due to the above-mentioned advantages,jellyfish search optimizer has been widely used to solve various complex optimization problems,but similar to other swarm intelligence algorithms,there are some disadvantages such as easy to fall into local optima,slow search speed and low search accuracy.Therefore,in this thesis,some improved versions of jellyfish search optimizer are proposed for multi-application scenarios,and the main work is as follows.(1)An improved jellyfish search optimizer for typical optimization problems.In order to cope with the shortcomings of jellyfish search optimizer,firstly,the Circle mapping is utilized to increase the population diversity and prevent premature convergence in the initialization phase.Secondly,the V-shaped transfer function is applied to convert the continuous space into binary space to solve the discrete optimization problem.In addition,the leader strategy is adopted to improve global search efficiency and solutions’ quality.Finally,the performance of the improved algorithm is verified by simulation experiments.(2)An improved jellyfish search optimizer for the prediction problem.Firstly,a new update mechanism of jellyfish search optimizer is designed,so that the update of the jellyfish search optimizer contains a wider range of stochastic solutions and the search capability of the algorithm is enhanced.Then,the improved jellyfish search optimizer is utilized to optimize the hyperparameters of long short-term memory neural network to complete the air quality prediction for Chengdu.Finally,the air quality prediction results show that long short-term memory neural network optimized by this algorithm has the higher prediction accuracy.(3)An improved jellyfish search optimizer for the diseases classification problem.Firstly,an improved time control mechanism is used to better balance the update process.Secondly,an adaptive function strategy is combined into the algorithm to improve the convergence speed of the algorithm.In addition,a superior population strategy is employed to improve the algorithm’s merit-seeking accuracy.Moreover,the jellyfish search optimizer with multiple improved strategies is used to optimize the weights and thresholds of the multilayer perceptron to complete the classification of breast cancer diseases.Finally,the corresponding datasets test results show that the multilayer perceptron optimized by the improved jellyfish optimization algorithm has the higher classification accuracy.In this thesis,the jellyfish search optimizer is improved to address its own strengths and weaknesses and is used for typical optimization problems,prediction problem and diseases classification problem.Experimental results show that the improved jellyfish search optimizer can be used for optimization problems in more application scenarios such as power dispatching,fault diagnosis and practical prediction problems.