Clustering Algorithm Of Data Streams Based On Ant Colony Algorithm

Data mining is a hot research area of applied mathematics. Recently, as the technology of computer science, communication and network develops rapidly, data streams, which grow at a rapid rate, reach continuously, evolve dynamically, have appeared in many areas. Clustering data streams is an important problem in data mining, a hot topic in recent researches as well. And features of data streams require higher standards on the algorithm of clustering analysis.A profound study is made in classic algorithms of clustering data streams in this paper, main work is as follows.Ant colony algorithm is applied into algorithm of clustering data streams, and an algorithm called Abclustream is proposed. This algorithm adopts two-tier framework, updates the micro-clusters by probability rather than distances, and method of probability is used to tell outliers from those are not. In addition, a detection mechanism is added to find out invalid clusters in time.Furthermore, since data steams with heterogeneous attributes exist in most occasions in real life, the algorithm proposed is promoted to analyze the cluster of data streams with heterogeneous attributes as well.Finally, the algorithm is compared with Clustream by doing experiments on computers, and the results imply that the algorithm in this paper has a better efficiency and quality. By this algorithm, it is easier to find out clusters of arbitrary shapes, better results is achieved for data streams with high dimensions as well.