Multi-database mining is an important topic for data mining. Database clustering is an effective method obtaining interesting patterns from multi-database. At present, some constructive related works for multi-database classification have been done, but there are many works need to do. In this paper, we propose a strategy clustering multi-database based on ROCK, which is a hierarchical clustering algorithm by using link. In our approach, we cluster databases in the light of the similarity of databases and the neighborhood information at the same time. We experiment our algorithm, and the results prove the rationality and efficiency of the method.
展开▼