Rough set based Privacy Preserving Attribute Reduction on Horizontally partitioned data and generation of Rules

Abstract: Data mining is the process of extracting knowledge from various databases. For security reasons organizations may partition their data horizontally or vertically. Another situation is hospitals maintaining sensitive information about patients. This can also be treated as horizontal partitioning of data if all the hospitals maintain data in the same format. Secure multiparty technique is a cryptographic method which can be used to derive useful results from partitioned data bases without violating the privacy of individuals or organizations. This paper describes a method for finding a rough set concept called reduct set from the partitioned databases using secure multiparty technique. From this reduct set, rules are generated using Naïve Bayesian algorithm. This paper also describes attribute reduction technique for horizontally partitioning databases .Generally it is necessary to implement data preprocessing operations on the data bases. This paper also describes an enhanced Transitive Closure algorithm for data cleaning.

Keywords: Rough sets;attribute reduction;data cleaning and rule generation