Mining Global Exceptional Rules in Multi-database

In multi-database there are four category patterns which refer to frequent itemsets or association rules. Exception rules have been defined as rules with low support and high confidence. Exceptional patterns reflect the individuality of branches and provide valuable knowledge about database patterns, so it is very important to make special policies for these branches. For multi-database mining, gaining global exceptional patterns from local patterns is the necessary process. In this paper, we mainly discuss the exceptional association rules mining. When mining exceptional rules in multi-database may be cause knowledge conflicts, we resolved these conflicts by correlation and designed an algorithm MGER-MDB. Finally uses the example to explain this algorithm.