Research On Parallel Ensemble Learning Of Glowworm Swarm Optimization And General Regression Neutral Network

The method of solving optimization problems,which involves many fields such as economics,finance,and engineering,has been a hot research topic.Swarm intelligence optimization algorithm is an important method for solving optimization problems,so it has attracted widespread attention.The design inspiration of most swarm intelligence optimization algorithm comes from natural laws and biological habits.Among them,the development of Glowworm swarm optimization(GSO)algorithm is based on the research of firefly's luminous behavior.GSO algorithm has the characteristics of simple and better search ability,and is widely used in many application fields.But GSO algorithm also has disadvantages,such as low accuracy of algorithm output,poor algorithm stability,and slow convergence speed.In this thesis,the theoretical foundation and research status of GSO algorithm are deeply researched.Considering its existing problems,different improvement strategies are proposed.And parallel ensemble learning of GSO algorithm and general regression neutral network(GRNN)is studied.The main research work is summarized as follows:(1)In order to solve the performance problems of the traditional GSO algorithm,this thesis uses a variety of strategies including mutation strategy,variable step strategy,and auxiliary location update strategy.And then this hybrid improved glowworm swarm optimization(HIGSO)is proposed,and a comparative simulation experiment was performed.The experimental results show that the HIGSO algorithm is superior to comparison algorithm in performance.(2)In order to improve the classification and prediction performance of GRNN,this thesis proposes a disturbance factor,and then the disturbance factor and smoothing factor are optimized by the HIGSO algorithm.In this thesis,HIGSO-GRNN parallel ensemble learning algorithm is proposed.Experimental results show the effectiveness and stability of the HIGSO-GRNN algorithm.(3)The prediction model is established,and the haze data of Beijing,Shanghai and Guangdong are collected.Then,HIGSO-GRNN parallel ensemble learning algorithm is applied to haze prediction experiment.Experimental results show that the method proposed in this thesis is feasibility and effectiveness in haze prediction.