- Share-frequent pattern mining discovers more useful
and realistic knowledge from database compared to
the traditional frequent pattern mining by considering the non-binary frequency values of items in transactions. Therefore, recently share-frequent pattern
mining problem becomes a very important research
issue in data mining and knowledge discovery. Existing algorithms of share-frequent pattern mining are
based on the level-wise candidate set generation-and-test methodology. As a result, they need several
database scans and generate-and-test a huge number
of candidate patterns. Moreover, their numbers of
database scans are dependent on the maximum length
of the candidate patterns. In this paper, we propose a
novel tree structure ShrFP-Tree (Share-frequent pattern tree) for share-frequent pattern mining. It exploits a pattern growth mining approach to avoid the
level-wise candidate set generation-and-test problem
and huge number of candidate generation. Its number
of database scans is totally independent of the maximum length of the candidate patterns. It needs maximum three database scans to calculate the complete
set of share-frequent patterns. Extensive performance
analyses show that our approach is very efficient for
share-frequent pattern mining and it outperforms the
existing most efficient algorithms.
(from crpit.com)
(local if available)
