Skip to Main Content
A methodology has been proposed for solving database problems requiring only approximate solutions. Data items are classified as sensitive and nonsensitive. An approximate transaction modifies only the nonsensitive data items which need not satisfy strong consistency constraints, and provides results only up to a degree of approximation. Further, it is shown that such an approach improves the performance in situations where transaction conflicts are frequent. Additionally, the methodology provides users as well as data managers with mechanisms to control the precision of the computation, preserving the qualitative characteristics of the data items.