Database View support for Predictive Intelligence
Summarize
Summary of Database View support for Predictive Intelligence
ServiceNow enables the use of database views as inputs for Machine Learning (ML) solutions within Predictive Intelligence. By leveraging database views, customers can join two or more tables into a single consolidated view, thereby expanding the range of fields and data available for training ML models. This approach enriches the solution outcomes by providing a broader dataset.
Show less
Key Features
- Support across Predictive Intelligence frameworks: Database views can be used in classification, similarity, clustering, and regression capabilities.
- Custom database views: Customers must create database views that join multiple tables relevant to their ML use case.
- Example use case: A similarity solution can utilize a database view combining multiple Knowledge Base tables (e.g., FAQs, How-To articles) to compare records across these joined tables instead of a single table.
- Record volume requirements: Each table within the database view must meet a minimum record count threshold (typically 10,000 records) to ensure meaningful ML results. If tables do not meet this threshold, related solution outputs may be incomplete.
- Configuration: In similarity solution definitions, the database view replaces the single table selection, allowing the ML model to process and compare data across multiple tables simultaneously.
Practical Benefits
- Expands training data diversity and volume by combining multiple related tables.
- Improves the accuracy and relevance of ML models by incorporating richer datasets.
- Enables more comprehensive similarity comparisons and other ML analyses across joined data sets.
- Provides flexibility to tailor ML inputs to complex data structures without duplicating or manually merging data.
Considerations
- Ensure database views are properly created and validated before associating them with ML solutions.
- Monitor and maintain the minimum record count per table within the view to avoid incomplete or missing results.
- Contact ServiceNow Customer Support to adjust minimum record thresholds if needed.
Use database views to join two or more tables as input for your Machine Learning (ML) solution.
Using database views in an ML solution
Database views help expand the amount of fields your solution can use for training. By using more than one table in your ML solution definition form, you can access more input data to help enrich the solution outcome.
Database views enable you to join two or more tables into one consolidated view. For this to work in an ML solution you must first create the database view. See Database views.
Database views are supported in all four of the Predictive Intelligence capability frameworks: classification, similarity, clustering, and regression.
A database view example for Predictive Intelligence
In the following example scenario, you've created a database view for use in an ML similarity solution. The image below shows the database view record you've created, including its Name and Label.
When you click the record Name, its content appears, as shown in the image below. Within the database view content, the five Knowledge Base tables you've joined to the view are listed. Most of these tables contain different Knowledge article template types, such as an FAQ or a How To article.
When a database view is used as input to a similarity solution, each of the tables that constitute the view must have at least the required number of records set in the configuration of your ServiceNow instance. The default minimum number required is 10,000. For example, the Knowledge View database view has five tables and each table must have 10,000 records. If a table doesn't have 10,000 records, you may not see the results from that table. If you must change that value, contact Customer Support.
In the image below you can see the similarity solution definition record you've already created, which you plan to associate to your database view. When you click the Label for your similarity record, its Similarity Definition form appears.
Similarity Definition forms compare your existing table records based on their similarity by using a table in the Table field and another table in the Test Table field.
To use a database view in your similarity solution, instead of selecting a table in the Table field, you select the database view you created, which in this example scenario is the Knowledge View database view. This configuration increases the number of records your solution uses in training because the system compares and processes five tables of data instead of one.