Research And Application Of The Dragonfly Algorithm Based On Enhancing The Exchange Of Individuals' Information

The optimization of swarm intelligence algorithm has always been a hot topic in the field of science and engineering research.The bionic mechanism and heuristic ideas entailed in it have made the solution to engineering mathematics optimization problems very quick and elegant.The dragonfly algorithm(DA)is a kind of relatively novel intelligence algorithm with few control parameters and simple principles,so it can be better applied to function optimization.This paper studies the dragonfly algorithm carefully,and analyzes and discusses the background significance,research status,source of inspiration,algorithm design and implementation,and simulation results of the basic dragonfly algorithm.It also gives the convergence proof,persudo code and flow chart of the dragonfly algorithm.Then,based on the basic theoretical steps and running results of the dragonfly algorithm,the advantages of the algorithm are analyzed and the disadvantages of the algorithm are also pointed out.On this basis above,three strategies(greed,equilibrium,and combination)are used to improve the dragonfly algorithm,and the dragonfly algorithm based on enhancing the exchange of individuals' information(EIDA)is proposed.The greedy strategy is the reservation of the historial optimal solution of the population.The equilibrium strategy is the optimization of the global search transmitting to local development.The combined strategy is the improvement of the population location update.This paper describes in detail the parameter selection,execution steps and flow chart of EIDA,and analyzes the population diversity and convergence of EIDA.Next,this paper uses seven benchmark functions to test the DA,EIDA,ABC and PSO.Four evaluation indexes,namely the optimal solution,the worst solution,the average solution,and the mean square error,are used to evaluate the performance of these four swarm intelligence algorithms.The average convergence curve for the four algorithms is also shown.Simulation tests show that EIDA can effectively improve the performance of the basic DA and improve the optimization ability of the basic DA.Inparticular,the optimization results of the high-dimensional multi-peak function show that the EIDA has a faster convergence rate,a higher search accuracy,and stronger resistance to local optimal solutions.Finally,the shortcomings of traditional time series forecasting methods are explained,and the combination of swarm intelligence algorithm and neural network is attempted to improve the accuracy of time series forecasting.This paper attempts to organically combine the two single algorithms,EIDA and Elman neural network,and proposes the EIDA-Elman time series prediction model to maximize the advantages of both.The introduction of EIDA is to solve the problem that the weights and thresholds of Elman neural network can easily fall into a local optimal solution during training.Based on the detailed description of the principle and algorithm steps of the EIDA-Elman model,it is applied to the Weibo hot topic prediction problem and performs simulation tests.The EIDA-Elman prediction model is evaluated by using two commonly used indicators: mean square error and relative error.The test results on Weibo topics show that the EIDA-Elman predictive model has good training conditions and high prediction accuracy.