Ant Colony Algorithm For Improvement And Its Application In The Cluster Analysis

Ant colony algorithm is a kind of simulated evolutionary algorithm based on bionic principles which simulates ants'behavior of finding foods. The algorithm has several virtues such as distributed computing, positive feedback mechanism and heuristic search ability, and has demonstrated its great potentials for development in solving many complex optimization problems. It was proposed by M. Dorigo in 1991, and was successfully applied in several areas. Ant colony algorithm is used for data mining in recent years, which provide a broader line of thought for clustering analysis of data mining.In this paper summarize the current situation of ant colony algorithm, and have research and analysis on it, then an improved algorithm is proposed. The main work is as follows.(1) The study of ant colony pheromone increment. This paper summarizes the two typical improved ant colony algorithm——Ant Colony System(ACS) and Max-Min Ant System(MMAS). Due to the two improved algorithm for lack of pheromone update rules, the pheromone increment dynamic hybrid update strategy is proposed. It can effectively use the global optimal solution and iterative optimal solution combined, and it can overcome local optimization problem in further, which is beneficial to find more optimal solution. In addition, it combined with mutation operation improving the whole group performance and reducing computing time.(2) The study of the single Swarm Intelligence. The paper firstly introduces the principle of clustering analysis, digital model and some usual method of clustering classification and the standard of clustering evaluation. Because the traditional clustering algorithms are limited to solve complex problems, it studies several bionic intelligent clustering algorithm, then the basic idea and the implementation process of clustering based on genetic algorithm, the ant colony algorithm and particle swarm algorithm is proposed. It lay the foundation for theoretical groundwork and Simulati- on contrast experiment with the improved algorithm proposed in fourth chapter.(3) The study of based on hybrid crossover operator of ant colony clustering algorithm. In order to make full use of several single Swarm Intelligence Clustering advantages to form a complementary strength, the paper analyzes several representa- tive hybrids clustering algorithm in few years. A new hybrid crossover operator of ant colony clustering is proposed. The algorithm combined with the advantage of ant colony algorithm and genetic algorithm, adopts the improved hybrid crossover operator. Two control mechanisms—the well-phased control strategy, heuristic multipoint crossover strategy are built up a hybrid crossover operator, the well-phased control strategy dynamically adjust the crossover scale, which significantly reduced crossover operation of invalid probability, heuristic strategy is built on the basis of the fitness can be effectively retain father generation good genes, thus the crossover operation can get unknown space exploration and known regional refinement.The paper studies and improves the ant colony algorithm of pheromone increment and space search capability. The result shows the proposed algorithm has improved the convergence speed,the clustering results and the robust of the ant colony algorithm.It provides certain theoretical foundation for the ant colony algorithm, and it has certain practical significance.