P2DM-RGCD: PPDM Centric Classification Rule Generation Scheme
In present day applications the approach of data mining and associated privacy preservation plays a significant role for ensuring optimal mining function. The approach of privacy preserving data mining (PPDM) emphasizes on ensuring security of private information of the participants. On the contrary majority of present mining applications employ the vertically partitioned data for mining utilities. In such scenario when the overall rule is divided among participants, some of the parties remain with fewer rules sets and thus the classification accuracy achieved by them always remain questionable. On the other hand, the consideration of private information associated with any part will violate the approach of PPDM. Therefore, in order to eliminate such situations and to provide a facility of rule regeneration in this paper, a highly robust and efficient rule regeneration scheme has been proposed ensures optimal classification accuracy without using any critical user information for rule generation. The proposed system developed a rule generation function called cumulative dot product (P2DM-RGCD) rule regeneration scheme. The developed algorithm generates two possible optimal rule generation and update functions based on cumulative updates and dot product. The proposed system has exhibited optimal response in terms of higher classification accuracy, minimum information loss and optimal training efficiency.
