Identifying Interesting Patterns in Multidatabases

Abstract

In this chapter we develop a new technique for mining multidatabases. The new mining algorithm, by comparing to traditional multidatabase mining strategies that have been focused on identifying mono-database-mining-like patterns, is able to identify both the commonality and individuality among the local patterns in branches within a company. While the commonality is important in terms of global decision-making, exceptional patterns often present as more glamorous than commonality patterns in such areas as marketing, science discovery, and information safety. We evaluated the proposed technique, and our experimental results demonstrate that the approach is efficient and promising.