Knowledge discovery in multiple databases (Q1887539)

The book contains the latest on research in database multi-minig (32 papers published after 2000) and offers for consideration a local-pattern analysis framework for pattern discovery from multiple data sources. Starting from the local pattern in multiple data bases, the authors propose, firstly, a new pattern named ``high-vote pattern'' based on statistical analysis of vote ratio received by a pattern from each branch of the company. Secondly a new pattern is introduced, referred to as the ``exceptional pattern'', which is illustrated with an algorithm for finding high-vote patterns in multiple databases. The first metric measures the relationship between the voting rate ``voting(ri)'' and the average voting rate ``AverageVR'', respectively, \[ \text{EPI(ri)}=(\text{AverageVR} -\text{voting(ri))/AverageVR.} \] Similarly, the second metric is defined using the support of the pattern in a branch of the company. Besides introducing the ``high-vote pattern'' and the ``exceptional pattern'' that improve the performance of multi-database mining systems and add to the automation of these systems, the authors also consider a fuzzy logic controller for providing a good man-machine interface in multi-database mining. Contents of the book: 1. Importance of multi-database mining; 2. Data mining and multi-data base mining; 3. Local pattern analysis; 4. Identifying quality knowledge; 5. Database clustering; 6. Dealing with inconsistency; 7. Identifying high-vote patterns; 8. Identifying exceptional patterns; 9. Synthesizing local patterns by weighting; 10. Conclusion and future work.