Modification And Application Of Particle Swarm Optimization Algorithm Based On Biological Behavior Mechanism

Particle Swarm Optimization (PSO) is a typical swarm intelligence optimization algorithm.Its basic idea stems from the simulation of biological acts in community-based social. Theexternal information is not used in search process. The fitness function value is the basiscriterion of evolution. It is a kind of intelligent algorithm character as "generate and test".PSO has been successfully applied in many practical problems of engineering optimization.As the origins of the PSO simulation community types from certain acts of gregariouscreatures, there are some difficults hard to overcome. It is a potential way to improve thereliability and effectiveness of PSO by using bio-behavioral mechanisms. This articleproposed several new improved PSO algorithms based on the perspective of biologicalbehavior mechanism. The improved algorithms are applied to the optimization and predictionproblems in the management practice to expand the application field of particle swarmoptimization algorithm.The dissertation combines together theory study and application analysis. The maincontents of this dissertation are as follows:(1) The problems and the improved way of PSO. This paper analyzed the basic idea,classification and characteristics of the bionic optimization algorithm systematically. Thebasic principles, main methods and development trends of swarm intelligence algorithm aresummarized according to the author’s understanding. The basic principles, existing problemsand its reasons of particle swarm optimization are analyzed to provide ideas for exploring theimproved ways of swarm optimization algorithm.(2) The PSO based on bacterial quorum sensing mechanism. The related researches ofembedding the biological behavior into PSO to improve the performance of algorithm areanalyzed. The PSO based on quorum sensing is proposed by combining the bacterial quorumsensing mechanism with PSO. The improvement is verified by six criteria simulation testfunctions. The different groups of quorum sensing frequency are tested for PSOQS to get thebest time which quorum sensing occurs. As the size of swarm will affect the success rate ofintelligence algorithm, we test different population sizes on PSOQS in the best quorumsensing frequency and analyze the results of the test.(3) The PSO based on parasitic immune mechanism. The behavior of bacterial parasiteand the mechanism of parasite immune are analysised in this part. We proposed PSO based onparasitic immune mechanism (PSOPI), which embedded the the basic ideas of parasiteimmune mechanism. The parasite group has elite group learning mechanisms to improve the ability to escape from the local extrema. The host group will acquire immunity after parasiticbehavior occurs in order to enhance the diversity of the particle in host populations. Theperformance of PSOPI, CPSO and PSOPB is analysis comparatively by simulating withstandard test functions.(4) PSO based on immune escape mechanisms. We explorate the immune evasionstrategies and mechanisms according to the phenomenon host immune response against theparasite. The bacterial immune evasion mechanisms are embedded into PSO. We adoptGaussian variation (Gaussian) and Cauchy mutation (Cauchy) to simulate the parasiteimmune avoidance behavior, form the PSO model based on immune escape mechanisms. Thecorresponding experimental parameters are set for standard test functions simulation.(5) PSO based on microorganism foraging mechanism. After summarize the law ofbiological foraging behavior, we explore the nature of biological foraging mechanisms. ThePSO based on bacterial foraging mechanism (PSOBF) is constructed by embedding thebacterial foraging behavior. The PSO based on honeybee foraging mechanism (PSOHF) isconstructed by embedding the honeybee foraging behavior. The effectiveness of the improvedalgorithms is tested by the standard test functions.(6) The practical application of the improved PSO. This article construct a distributioncenter selection model to test the optimize capability of the improved PSO. The result showsthe good performance of the improved algorithm. To test the forecast capability of theimproved PSO, we train BP neural network with PSOPI and get the best tray of weights forthe Shanghai stock market price index prediction. The results show the effectiveness of thethe improved PSO algorithm.