Research On Improved PSO Alcorithm Based On Similarity Of Variation And Its Application

The emergence of population-based intelligent optimization get rid of thedilemma which the traditional optimization technology face to, as it not only providesa new ideas for optimization problems solving, but also solves many practicalproblems including the fields of engineering, chemical engineering, economics, andneural networks etc. Particle Swarm Optimization algorithms (Particle SwarmOptimization,PSO) is an adaptive stochastic optimization algorithm, featuring simpleprinciple, less adjustment parameters, faster convergence, presenting an enormouspotential in practical engineering application.However, PSO, as a new algorithm, has not reached its full maturity in the aspectsof theoretical analysis, and practice application area are is not full mature, andparticles group algorithm itself also exists limitations. The thesis conducts a deeperstudy from these shortcomings, including the main research contents as follows:1． The author recognizes that the reasons for a shortage of diversity in the laterstage algorithm is that particles is easy to be convergent with the optimal particle inpopulation, leading to the loss of local search ability of particles. Based on this, thethesis proposes an improved algorithm, which defines the concept of aggregation todetect the diversity of population, based on Euclidean distance among particles. Thelost of population diversity, namely the measurement of similar degree betweenarticles and global optimization particles is found. The author defines optically thenew variation strategy to particle variation, based on similarity, so as to enhance itslocal search capabilities. The author emulates the improved algorithms thoughMATLAB and compares it with several improved classic algorithms, withexperimental results showing the efficacy and superiority of the improved algorithm.2．The thesis adopts two optimization problems in e-commerce domain: thepricing problem of seasonal products and the site problem of logistics distributioncenter. First, the author makes a introduction on the related knowledge of this twoproblems, and conducts a solution to the model according to the case at last. On onehand, the author confirms the effectiveness of the group algorithm of improvedparticles proposed in the thesis; on the other hand, the thesis widens the needed rangeof particles group algorithm.