Frequent Itemset Mining of Multiple Database s

Oftenti mes we need to investigate m ore than one source of data to provide a solution to the proble m at hand. This data integration proble m has been investigated and largely solved for simple situations in traditional relational database m a n age me nt syste ms (RDBMSes). They typically provide a m e a ns for the user to join datasets together based on a co m mo n si mple attribute. Not all attributes are simple however, and m uch work re m ains to be done to opti mize queries for m ore co mplex criteria across m ultiple datasets. This paper describes several new and effective techniques for opti mizing data mining queries in such situations. Practical solutions to proble ms such as query representation and query opti mization are presented, followed by results of recent experi me nts w hich de monstrate improve me nts over previously k nown algorith ms.