Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Automated basic theory > Artificial intelligence theory

Research on Clustering Algorithm Based on Mutation Particle Swarm Optimization

Author WangDong
Tutor LuoKe
School Changsha University of Science and Technology
Course Applied Computer Technology
Keywords Data Mining Particle Swarm K - means clustering algorithm Variation Clustering
CLC TP18
Type Master's thesis
Year 2011
Downloads 17
Quotes 0
Download Dissertation

Most work of data mining has focused on the discovery of the methods which could effective cluster to large database. At present,there is a large number of clustering algorithms in which k-means algorithm was applied widely.K-means clustering algorithm regard point as the prototype for clustering of spherical data.Thoughts of the algorithm is simple,easy to implement,fast running speed,small memory consumption and handling large data sets,but there are some major disadvantages:only then in the situation of definite the starting value,the cluster result is the only ascertained;The algorithm is the partial searching optimization algorithm,easy to fall into the partial minimum for tracking down the objective function.Moreover,the algorithm relies on the initial classified choice on a great extent.If the classification seriously deviates the overall superior calssification,the algorithm very possibly falls into the partial minimum and obtains a partial optimal solution.On the other hand,the structure of the PSO is simple and the very quick running rate,so the PSO algorithm is used to the cluster algorithm. On the basis of previous theory,the algorithm was improved on this paper and the two algorithm were combined organicly.Work as follows:1. The clustering was completed with the variation PSO. First ,analyzing the shortcomings of the particle swarm algorithm and variation of the particle is introduction to PSO,the premature convergence phenomenon was overcomed by increasing the diversity of the population.Secondly,improving the algorithm’s accuracy and convergence speed through the adjustment of the inertia weight.Finally,the K-means algorithm and pso was combined to a hybrid clustering algorithm.the algorithm effectively balance the exploration and development of the pso in the process of optimization,thus ensuring the stability and convergencing to the global optimum of the pso.2. The clustering was realized by the pso clustering algorithm based on population diversity.In the first place,analyzing the shortcomings of the indicators of population diversity.In the second place,the mutation of the pso and K-means algorithm were introduced to the pso;Finally,the particles was appropriate disturbance by the spatial characteristics.Not only is the pso local search ability improved,avoiding the premature convergence of the algorithm by increasing the population diversity.3.computer simulation.simulating of the proposed algorithm was implemented by using the VC-6.0 tool, and compared the proposed algorithm with the existing results,and analysising performance of the algorithm.

Related Dissertations
More Dissertations