DP-FIM FOR DATA LINKAGE USING ONE CLASS CLUSTERING TREE FOR MOVIE RECOMMENDATION

Abstract

Frequent itemset plays a vital role in many datamining fields such as- finance and biology. However, release of powerful patterns and trends are increasing concern on personal privacy, so the problem is –How to perform frequent itemset mining on transaction database which satisfies differential privacy? The proposed approach is called as DP-FIM (Differentially Private Frequent Itemset Mining) which can concurrently provide huge level of data utility and huge level of data privacy. It is very difficult to achieve data utility and privacy when existing system having long transaction. A system uses a transaction splitting technique to divide a long transaction into sub-transaction whose cardinality is no more than specified limit. To avert information loss generated by transaction splitting, run time estimation method is used to estimate defined support of itemset in original database. Data linkage compares records in one database with records in another database to match them. Data linkage is performed among the tables to cluster the data. But, existing data linkage methods do not handle one to many, many to many and many to one linkage field values. This system investigates a method which links different entities by using OCC-Tree (One Class Clustering Tree) in which inner node contains attribute from one database and leaf node holds close representation of matching records from another database.