Research On Improvement And Application Of Multidimensional Optimization Based On Particle Swarm Optimization

As a kind of swarm intelligence algorithm,particle swarm optimization algorithm is an algorithm that simulates the foraging behavior of birds.Particle swarm algorithm is favored by scholars because of its simple implementation and few parameters.However,it still has the shortcomings that it is easy to fall into local optimization and cannot solve large-scale high-dimensional optimization problems.In order to solve large-scale highdimensional optimization problems,prevent falling into local optima,and improve the performance of particle swarm optimization,scholars have improved and optimized the algorithm from multiple dimensions.Parameter improvement and optimization method improvement are the most commonly used improvement methods by scholars,but often because the improved strategy does not take into account the balance of global search and local search,the algorithm can only exert performance on specific problems.Therefore,this paper first proposes to divide the algorithm iteration process into early and late stages,design an inertia weight attenuation strategy and an adaptive mutation optimization strategy,and complete the improvement of the particle swarm algorithm.The main research work is as follows:1.Propose a particle swarm optimization algorithm based on normal distribution decay inertial weights,using normal distribution curves to make non-linear changes to the inertial weight parameters,so that the inertial weights are in a larger state in the early stage and continue a certain iterative process,and the overall maximize search capabilities.From the early stage to the later stage,the inertial weight decays sharply.After the transition to the later stage,the inertial weight is in a small state,which maximizes the local development capability.In the algorithm simulation experiment,it is verified that the strategy takes into account the balance of global search and local development from the perspective of parameter setting.2.Propose adaptive mutation particle swarm optimization algorithm,design Cauchy mutation operator and Gaussian mutation operator,design new particle velocity update formula combined with iterative process,and integrate the strategy of attenuating inertia weight based on normal distribution.From the perspective of improving the optimization method,the global search and local development capabilities are balanced.In the algorithm simulation experiment,the 500-dimensional test function is used for highdimensional testing.The adaptive particle swarm optimization algorithm has excellent performance.The optimal solution is found in some high-dimensional test functions.The convergence accuracy and convergence speed of other test functions far exceed the comparison algorithm.3.The adaptive mutation particle swarm optimization algorithm based on the normal distribution decay inertia weight strategy will be integrated for application analysis,and a four-input-four-hidden-layer BP neural network model will be designed to predict the number of COVID-19 diagnosed daily,using adaptive mutation Particle swarm optimization algorithm to optimize the weight and threshold parameters of the BP neural network model.Collecting the data of the daily number of newly diagnosed people from the National Health Commission of the People's Republic of China and the US Department of Health,conduct simulation experiments,and analyze the fitness value and the performance of the test set.From the simulation results,the performance of the adaptive mutation particle swarm optimization algorithm is excellent,and it achieves the effect of improving the accuracy of the model.