Research Of Clustering Analysis Based On Using Ant Colony Algorithm

The fast development of information and network technology causes the phenomenon of"Information exploding but knowledge poor". It becomes more and more tough day by day. In order to solve this problem, data mining was put forward. As a new technology, data mining grew up vigorously and fast in the specific environment. And it demonstrated its strong vitality in different circumstances. In some primary websites, data mining was chosen as one of the most popular technologies in the future. Nowadays, more and more companies especially super ones such as IBM and Microsoft devoted many resources to diving into data mining. Data mining is a technology of multiple disciplines. It is involved in many subjects such as database, artificial intelligence, statistics, knowledge-discovery, biology-computing and so on. The development of data mining would affect the process of global information greatly. So it is necessary to dive into data mining systematically, roundly and deeply. This dissertation studied and analyzed data mining deeply, especially the clustering analysis, which takes a great part of data mining researches. And some ideas and improvements are proposed afterward. The main contents of this dissertation are listed as follows:Description of clustering analysis in data mining. The concepts of data mining and data mining system are introduced at the beginning. Then the classification, process and main tasks of data mining are introduced briefly. As one of the most important portions in the research of data mining, clustering analysis is mainly used to discover the valuable data distribution and data mode in the potential data. Clustering analysis, including its definition, methods, data types and the standards of measuring the results are presented as well.Description of basic Ant Colony Algorithm (ACA). The intelligent colony algorithms, which are abstracted from natural biologic colonies, are only a slice of bionic algorithms that generated by long term observation. As a representative bionic algorithm, ACA shows great performance and tremendous potential of evolution in solving complicated optimization problems, especially discrete optimization problems. In this dissertation, basic ACA is described from its biological theories and characteristics of system to its mathematical model and implementation. At last, the temporal and spatial complexities are analyzed as well. Clustering combination method based on improving ACA. On the basis of learning basic ant colony clustering model, LF algorithm and LF algorithm based on entropy, Single Ant-colony Clustering Algorithm (SACA) is brought forward. Then Multiple Ant-Colonies Clustering Combination Algorithm (MACCA) is introduced as well. Its main theory is that after parallelized running several SACA at different velocity styles, the results are transformed to a hypergraph by the hypergraph model, and at last the hypergraph is compartmentalized by graph-division algorithm based on ant colony algorithm. The MACCA test results of functionalities and performance are present as well.