Dependencies - trillium_discovery - Latest

Trillium Discovery Center

Product type
Software
Portfolio
Verify
Product family
Trillium
Product
Trillium > Trillium Discovery
Version
Latest
Language
English
Product name
Trillium Discovery
Title
Trillium Discovery Center
Copyright
2024
First publish date
2008
Last updated
2024-10-18
Published on
2024-10-18T14:55:05.094442

A dependency is a many-to-one data relationship where one or more attributes determine the value of another attribute within a single data source. After data is imported into a repository, review the results by examining each potential dependency found.

There are two types of dependencies:
  • Discovered—Found as a potential dependency but not yet reviewed for validity.
  • Permanent—Reviewed and verified.

A discovered dependency is considered a potential dependency until you validate it as permanent.

Note: Dynamic data sources do not support dependency analysis. Dependency analysis is available for profiled (fully-loaded) data sources only.

How Dependency Analysis Works

When you import data (by creating a data source), the Discovery Center automatically analyzes the data and identifies potential (discovered) dependencies on a sample of 10,000 data rows. This process discovers single- and double-attribute dependencies that are at least 98% consistent and does not include attributes that are less than 2% unique.

To find a larger number of potential dependencies, you can run a dependency analysis using criteria that allow you to find dependencies that have:

  • Less than 98% consistency. This is determined by the Quality % metadata. Quality % is the measure, as a percentage, of how good (consistent) the dependency is. Dependencies listed with a quality of 100% represent dependencies within the data with no conflicts. Dependencies listed with a quality of less than 100% represent dependencies containing conflicts. For example, if there was a dependency between the attributes Order Number and Order Date and the quality percentage was 99%, this indicates that 99% of the time the same Order Number will have the same Order Date.
  • More than two combined attributes (fields/columns).
  • Attributes less than 2% unique.