Research And Application Of Clustering Analysis Method Based On Ant Colony Algorithm

The phenomenon about"the data explode but knowledge is poor"caused by rapid development of information technology and network technology is serious day by day. Under this environment, data mining was put forward as a kind of new technology, which demonstrated its strong vitality more and more.In some domestic primary website,data mining technology was chosen one of the most popular technologies in the future. Data mining is a research field of crossing multi-disciplines, and its development will influence the process of the global informaionization greatly. So it is the desirability of informationization development to have a research on data mining technology in an all-round and systematical way. Clustering analysis as one of data mining analysis methods is the performance is prominent, including the pattern recognition, the data analysis, the imagery processing, as well as the marketing research and so on many domains obtain the widespread application.This dissertation deeply studies and analyses the data mining technology, especially the one for clustering analysis. And then some ideas and improvements have been proposed. The main contents are listed as follows:Description of the data mining technology in brief. The clustering is one kind of important data analysis technology, which is an essential part in the data mining theory frame, through searches and distinguishes the data union, the description data, discovered between the data conceals and the interesting mutual relation.The basic concepts of data mining technology are introduced first. Based on the basic concepts of data mining, this dissertation classifies and summarizes the objects of data mining, the findable patterns , common method and application in detail.Description of clustering analysis in brief.The definition, data type, primary algorithms and mesaure standard are briefly introduced. All of the above become the basis for this dissertation.Description of the ant colony algorithm(ACA) in brief. ACA is a new-type bionic algorithm which comes from the resarch on the biological world of nature. It is a typical representative algorithm of Swarm Intelligence which have good performance for discrete space problem. This article draws out the ant group algorithm by the community intelligence, and explained the principle and characteristic of the algorithm in detail. Through the analysis ant colony algorithm good and bad points, point out this algorithm an improvement research direction. In view of the ant group algorithm precocious, the stagnation and easy to fall into partially the most excellent flaw, the thought fuse which with the aid of the ant group algorithm and the genetic algorithm ,which can solute separate territory question to the more effective solution continual territory question transition. Improved the overlapping variation operation ant group algorithm, which can enhance the ant group algorithm the performance.Enhances the performance the ant cybotaxis class combination algorithm research. Research on the algorithm on ACA, an improved LF algorihm is presented firstly, through change formula, radius, spatial division and with the aid of short memory. It has reduced the number of the parameters and the computation time. Unifies the K-means algorithm, an improved K-means algorithm based on the pheromone is introduced detailedly. The algorithm works with the transformation probability to realize the clustering. It has reduced the number of the parameters and improved the speed of clustering. According to two algorithm characteristics, an algorithm based on ACA and clustering algorithm combination is proposed, which include two clustering courses. The first is to accomplish the task of clustering by using ILF algorithm, the datum are collected and identified with the method SACA. At last, the result of clustering will be ptimized by the algorithm of K-means based on the pheromone inorder to get more superior performance compare to single algorithm.At last,combination algorithm is tested by some datasets and used to analyse bank credit card customer spending behavior.