The Research Of Fuzzy Cluster Methods Based On Immune Theory

For the problem that the classical clustering algorithm is usually sensitive to initial value or easy to bring about local optima,a novel clustering algorithm is provided which is based on model of C-means and the artificial immune cells.Namely,we imitate an immune principle to execute cluster.The principle is that immune cells first change into mature cells,and then polarize into antibodies or memory cells.Importantly,the new arithmetic put forward a double selection method based on affinity between cells to ensure selected cells more typical;bring into a self-adapting mutation method to accelerate speed of global convergence;select the best antibody group according to antibody inhibition principle;use immune memory mechanism keep good cells and give away bad cells.In Simulation,we make a comparison between the new arithmetic and the two algorithms:Genetic Guidance Arithmetic(GGA)and Artificial Immune C-means Arithmetic.The result shows that new arithmetic appears several features such as high accuracy of clustering and better clustering capabilityBased on the above-mentioned,this paper researches the clustering algorithm based on particle swarm optimization algorithm.The immune information evolutionary mechanism of Artificial Immune System is used into Particle Swarm Optimization Algorithm(PSO),a new dynamic clustering algorithm based on C-means and improved PSO is presented and it can avoid "early ripe" of PSO and traditional clustering algorithm.The particle sufficiency can judge the clustering scheme it corresponding to,and then we can select parts of particles into multiple-point mutation.We can get the initial clustering numbers according to the experiential rule k_max≤n^1/2(n is the number of sample)of classical clustering theory and find the best k value through swarm performance cost function.Simulation shows that the combination of PSO has high global convergence ability and the immune evolutionary mechanism has high local convergence ability can efficiently raise their respective handing problem ability.The new arithmetic can avoid "early ripe" and has higher accuracy of clustering than clustering only based on PSO.Through theory analysis and experiment proof,we can obtain that immune idea can improve cluster quality effectively.The new cluster algorithm proposed in this paper is useful and valuable for practical cluster problems.