Many large organizations need to mine multi-databases distributed in their branches for exceptional pattern for the purpose of globally decision-making. The present major strategies of mining exceptional interesting pattern is to merge all multi-databases into a single dataset for discovery, but this destructs the local distribution character of the pattern in different branches. The only work mining multi-database not as a single database is not complete and the method to find exceptional patterns is inaccuracy. In this paper, we give a new method to mining exceptional interesting pattern in multi-database. The experimental results show that our theory is practical and efficient.
展开▼
