Design complexity has significantly increased for deep submicron technologies. It is very difficult to design, simulate and verify the whole system without decomposing it into smaller subsystems and rebuilding the database bottom-up based on clustered optimized data. Especially for the analog applications, most of the front and back-end partitioning techniques solve the hierarchical tree graph based on geometrical aspects without gleaning information from the functional aspect. This work proposes an electrically aware database construction scheme that adopts a signature-based statistical partitioning technique. The proposed technique aims primarily at removing redundancies and improving the database construction accuracy and runtime.