Db2 for z/OS version 13 includes a real gem of a capability that few people know about. This is the built-in capability, unique to version 13, called SQL Data Insights. SQL DI allows DBAs to embed deep learning and machine learning capabilities into your existing Db2 tables and views. It combines artificial intelligence (AI) with advanced IBM® Z technologies to infuse the Db2 engine with SQL-based semantic queries on user tables and views.
Semantic queries are requests for items in a database that have a relationship to other items, based on the meaning of the item. For example, suppose you wanted to query a Db2 table looking for people who are similar to other people based on a number of factors. Consider a bank that would like to better understand how some of their existing customers might be good prospects for a Wealth Management Practice they have created. With SQL DI users can query for customers who are similar to the best current Wealth Management Practice customers. This can be determined with SQL Data Insights using a built-in function called AI_SEMANTIC_CLUSTER. In SQL Data Insights a query can show which potential customers are semantically similar to the best existing customer.
Or consider how a telecommunications company might want to know which customers might leave their service by tracking those who have already left. The company could use the SQL DI function AI_SEMANTIC_CLUSTER or AI_SEMANTIC_SIMILARITY to find existing customers who are most like the group of customers who have already left the service. The company could then concentrate their outreach on special offers for those vulnerable to quitting to entice them to stay.
The traditional functionality associated with a relational database system is not capable of capturing the semantic relationships among value entities within and across columns or rows. SQL DI uses database embedding, a self-supervised learning approach in deep learning and AI, to train a neural network model and infer semantic meanings for the unique values in a relational table. The inferred meanings in the form of numeric vectors encapsulate the inter-column and intra-column data relationships. SQL DI then uses the trained model that consists of numeric vectors to run AI queries that discover, match, and cluster semantic similarities and dissimilarities in your Db2 data.
Unlike most Deep Learning and AI tools, SQL Data Insights makes it easy to gain the same machine learning benefits, while using simple SQL queries. It's no longer necessary to set up a dedicated AI team with the latest deep learning skills and a PhD in AI to set up what can be a 6-9 month project to extract the data out of Db2, build complex models, test them, and then train users in how to use them.
Video demo here:
How to guide:
https://www.ibm.com/docs/en/SSEPEK_13.0.0/pdf/db2z_13_sqldibook.pdf
https://www.ibm.com/docs/en/db2-for-zos/13?topic=running-ai-queries-sql-data-insights