The Improvement Of The Ant Colony Algorithm And Its Using On Clustering Analysis

Ant Colony Algorithm is a new developed bionic optimization algorithm, which simulate the ants'finding food process. Because of its simple mechanism, plus-feedback parallelization, strong lustiness and excellent distributed computational mechanism, it has shown its excellent capability and huge developing potential in solving many complicated optimization problems. Despite the strict theoretical basis of ant colony algorithm has not yet been laid, and both at home and abroad related research is still in the pilot phase of the exploration and preliminary application, the ant colony algorithm's study had been from a single traveling salesman problem penetrated into the multi-application areas, from settlement of one-dimensional static optimization problem developed to solve the multi-dimensional and dynamic combinatorial optimization problems, and from discrete domain within the scope of research gradually extended to a continuous domain within the scope of researchHowever, there are still some faults in ant colony algorithm: as in performance, the diversity and stability of the solustions searched by ant colony are in conflict with convergence speed of the algorithm. This is because the individual movement of ant colony is random. Although through the exchange of information to the evolution toward the optimal path, they can hardly find a optimum one from a mass of paths within a short time when the problem scale is large enough, eyelessly accelerate the convergence speed will make the ants'local search and lead to premature and stagn ation of the algorithm easily.In this paper, we first introduce classical ant colony algorithlm and give a summary about its current research work. Through the research of the basic ant colony algorithm's shortcomings, we draw on the many advantages of artificial fish-swarm algorithm. Although there are some gaps of the artificial fish in determining the precise algorithm for the optimal solution, it can quickly converge to the optimal solution where the scope. According to this characteristic, we derive a degree of exciting concept, using it to express the extent of close to the optimal solution size. The greater the degree is, the closer to the optimal solution of the current solution will be, otherwise the farther away from the optimal solution. By determining the size of excited degree, we can quickly determine the scope of the optimal solution, so as to determine the accuracy of the second phase of the optimal solution and lay a good foundation. Through the simulation with the basic ant colony algorithm and multi-state comparison of ant colony algorithm to improve the conclusions drawn. At the last part of thesis, we give the ant colony algorithm in clustering applications, the focus of a new ant colony algorithm based on the improvement of K-means clustering algorithm, and give experimental results to improve.