Research Of Clustering Algorithm Based On AFSA With Application

Cluster analysis is an important data-mining technique, as well as an important step of data-mining. The task of clustering is dividing the object data into different sub-classes .The same sub-class of the object has a strong similarity and the different sub-classes was not similar to the object. The traditional clustering method is introduced in this paper as well as a new group of intelligent technology: Artificial Fish-swarm Algorithm. With the studies of artificial fish-swarm algorithm in clustering analysis, an improved Artificial Fish-Swarm Algorithm was proposed, which combined with k ? means algorithm.The basic artificial fish-swarm algorithm was de described in detail. An improved artificial fish-swarm algorithm (IAFSA) was proposed, which improved the parameters"step"and the behavior of fish—fish_prey. In IAFSA, the random step of artificial fish in the basic AFSA was changed into which depend on the food concentration differences between itself and Goal Artificial Fish. If the food concentration difference was more than evaluation function, the artificial fish would go ahead farther than random step. Meanwhile, in the fish _prey behavior description, when the maximum times of test to go ahead was tried and still could not find the status of the forward direction of improvement, the artificial fish would forward to the current best state in the bulletin board records as a certain probability.As we all know, the algorithm was sensitive to the initial clustering center in the k ? meansclustering algorithm. In order to solve the problem, a new KM ? AFSA algorithm based on the artificial fish-swarm algorithm was proposed. Many artificial fish was set in the clustering space even in the new algorithm, and then the improved artificial fish-swarm algorithm was implemented. The optimal artificial fish were initial cluster center in k ? means algorithm. A number of shortcomings of k ? means algorithm was improved by the combined of the Artificial Fish-Swarm Algorithm and k ? meansalgorithm.The common methods of regional economic analysis were instructed in the last of the paper. The application of clustering algorithm in regional economy analysis was studied .An economic analysis example used the KM ? AFSA algorithm was given .In the case of Shandong Province, the economic of 17cities of Shandong was analyzed by KM ? AFSA algorithm, which aimed to evaluate the economic station, improve and provide strong support of economic policies .