An uncertain database [1] is a kind of database studied in database theory. The goal of uncertain databases is to manage information on which there is some uncertainty. Uncertain databases make it possible to explicitly represent and manage uncertainty on the data, usually in a succinct way.
At the basis of uncertain databases is the notion of possible world. Specifically, a possible world of an uncertain database is a (certain) database which is one of the possible realizations of the uncertain database. A given uncertain database typically has more than one, and potentially infinitely many, possible worlds.
A formalism to represent uncertain databases then explains how to succinctly represent a set of possible worlds into one uncertain database.
Uncertain database models differ in how they represent and quantify these possible worlds:
Though mostly studied in the relational setting, uncertain database models can also be defined in other relational models such as graph databases [6] or XML databases.
The most common database model is the relational model. Multiple incomplete database models have been defined over the relational model, that form extensions to the relational algebra. These have been called [7] Imieliński–Lipski algebras:
The following table is a relation of an incomplete database, described in the formalism of NULL values:
| id | Name | Salary |
|---|---|---|
| 1 | Alice | 10,000 |
| 2 | Bob | NULL |
| 3 | Charlie | NULL |
There are infinitely many possible worlds for this incomplete database, obtained by replacing the "NULL" values with concrete values. For instance, the following relation is a possible world:
| id | Name | Salary |
|---|---|---|
| 1 | Alice | 10,000 |
| 2 | Bob | 8,000 |
| 3 | Charlie | 12,000 |