The Research On Clustering Algorithm Based On Bee Colony And Rough Set Theory

The explosive growth of data brings us into the real data and the information age. How to get potential and useful information from massive data to guide people to make the right decisions is inseparable from data mining. Cluster analysis has become an active research topic in data mining because of its unique ad vantages. But clustering algorithm has some shortcomings,so studing how to further improve the deficiencies of the clustering algorithm to solve the difficulties in social practice has very important significance.This paper is base on the traditional clustering algorithm.it discussed the shortage of clustering algorithms firstly,then discussed the colony algorithm,rough set and granular computing,finally,improved the shortcomings of the traditional clustering algorithms by combining with the improved colony algorithm,rough set and granular computing.1. Aiming to resolve the problems of the traditional K-means clustering algorithm such as random selecting of initial clustering centers, lacking the ability of handling boundary objects of data, the low efficiency, this paper brought in the granular computing and rough set and proposed an initialization method that was based on granular computing and max-min distance means,then effectively deal with the clustering problem of boundary data by rough set,finally used criterion function of equalization to get better clustering results.The results of experiments show that this algorithm has higher accuracy, less iteration times.2.Due to the disadvantages of traditional K-medoids clustering algorithm such as sensitivity to the initial selection of the center, the poor global search ability, instability, this paper proposed a K-medoids clustering algorithm based on improved artificial bee colony.Firstly,improved the shortcomings of the random selection of initial swarm and search step of traditional artificial bee colony,then further optimized K-medoids by the improved artificial bee colony algorithm,in order to enhance the performance of clustering algorithms. The results of experiments show that this algorithm can reduce the sensitive degree of the noise,has high accuracy and strong stability.3.An artificial bee colony rough clustering algorithm base on mutative precision search is proposed to improve K-means clustering with the problem of the poor global search ability and stability. The proposed algorithm generates initial swarm by density and distance, and gets the selection probability of onlooker bees according to the fitness and density of lead bees, then updates scout bees through the method of mutative precision search, in order to avoid falling into local optimum, finally, combines with rough set to optimize K-means. The results of experiments show that this algorithm not only can suppress effectively the local convergence and reduce the dependence on initial cluster center, but also has higher accuracy and stronger stability than others.4.A honey-bee mating optimization clustering algorithm is put forward with using the thought of honey-bee mating optimization algorithm. It generates initial swarm by density and distance, and regards rough set clustering algorithm as a code of the works to enhance the local search ability of the algorithm. At last, in order to improve the diversity level of the swarm and the global optimization ability of the algorithm, random swarm population are introduced continuously in the iterative process. Our experiments show that the proposed algorithm not only can effectively suppress premature convergence, but also has strong stability and produces good clustering results.