Research And Application On Ant Clustering Algorithm

Data mining is the process of automatically searching large volumes of data for patterns and rules. Cluster analysis, as one of the important research branch of data mining, has been widely applied to many other domains. It can be used as the tool of data mining to discover the in-deep information of data distribution in database, and also be considered as the pretreatment process of other data mining algorithms. As a new intelligent bionic algorithm, ant colony algorithm has showed good prospect in the cluster analysis because it has features of distributed parallel caleulation, information positive feedback and heuristic search ability.Based on fully studying the existing basic principle and characteristics of ant clustering algorithm. This paper proposed improved ant clustering algorithm which is based on ant pile of forming principle and fusion with an improved genetic algorithm, in order to further improve the algorithm convergence speed and cluster quality. Main idea of this algorithm is firstly generating initial distribution of pheromone for ant clustering algorithm using the rapid global search capability of the improved genetic algorithm. Then get the exact solution using positive feedback of ant clustering algorithm. This algorithm realized the two algorithms'strengths, greatly reduced the number of the testing by adjust parameters of the ant clustering algorithm. It also prevented the fault that the ant clustering algorithm quickly converge to a local optimal solution, made algorithm improved in optimizing performance and time performance. Finally, this algorithm is applied to customer segmentation of real estate industry.