The Research And Improvement Of Cluster Analysis Models Based On Ant Colony Algorithm

Ant colony clustering algorithms build clustering models through simulating the behaviors of the swarm intelligence of the ant colony, and they are also belonging to the cluster analysis algorithms which are based on models. In recent years, the research of ant colony clustering algorithms has become the hot spot and leading issue in the field of cluster analysis. As the studies finding, building the cluster analysis models to solve some problems is closer to actual clustering problems than the classic clustering algorithms.This paper focuses on the cluster analysis models based on the model that ant colony sort the young ants and the model of ant colony’s self-aggregation after studying a variety of ant colony clustering analysis models.On the basis of deeply studies to the LF algorithm, this paper proposes IDLF algorithm. It mainly contains the following five improvements:enhancing the moving purpose of the ants; adjusting the visual range of the ants dynamically; adjusting the adjustment coefficient of the group similarity dynamically; improving the termination condition of LF algorithm so that the algorithm can terminate according to the actual clustering automatically; increasing the solution to deal with the situation of long-term loading or unloading actions of the ants.AntTree algorithm model is one kind of the new types of ant colony clustering algorithm models. This paper researches this model and proposes some optimized solutions, it mainly contains:optimizing the initialization of the algorithm; optimizing the threshold update method; perfecting the moving rules of the ants.The experiments of both the improved algorithms are performed on three standard UCI datasets. The results of the experiments demonstrate that IDLF algorithm and the optimized AntTree algorithm enhance the clustering performance of the original algorithms.