Apply Machine Learning to Discover Rules in Existing Master Data
- Innovations
With SAP S/4HANA you can use rule mining to discover rules in existing master data:
- Use mining runs to analyze existing master data.
- Collaboratively decide on business relevancy of proposed rules from rule mining.
- Create and link data quality rule from accepted rules.
- Information from rule mining is used in implementation of data quality rule.
- Business value
- Ease and shorten the discovery of rules with machine learning.
- Efficiently qualify and implement discovered rules as data quality rules.
Master Data Rule Mining Process
The preceding figure illustrates the typical process in master data rule mining.
Mining Run
Mining Run guides the system on how to find interesting rules:
- Goal
- Set your expectation.
- Tables
A list of tables to be mined at the same time.
Focus Area: The data set you want to use for mining. For example, Product Type = Finished Goods (FERT).
Fields: Selected fields to find potential rules for:
- Checked by Rule: the THEN part of the rule.
- Condition of Rule: the IF part of the rule.
- Parameters
Maximum number of rules: top N best rules proposed by the system, by default 100
To start the mining run:
- You are informed about the volume of data for mining before you confirm the mining run start.
- A mining operation that is running can be stopped from the UI.
- The system executes a machine learning algorithm in the background job (asynchronous). The end user is free to leave the screen.
- To check the mining run status and the progress, check the message strip details in the header.
- A restart or edit of the mining run is possible when it is set to Error by the program or it is manually stopped.
Managing the Mining Run
- Filter by Status, Description, Focus Area Fields, or administration data, and so on.
- Navigate to the rules from rule mining.
- Create a new mining run by copying an old one.
- To delete the proposed rules with Initial status together, choose Delete Mining Run.
To explore a mining run:
- Navigate from a completed Mining Run to the rules of this run.
- Access the app from the launchpad to freely filter and search for rules from different mining runs.
- Search rules by Condition Field name and Condition Field Value, Status, and so on.
- Technical Description and Business View are both visible.
Review and Accept Rules
The figure illustrates the main facts about review and accepting rules.
Link to a Data Quality Rule
The figure above shows the possibility to create a link to a new quality rule. You can also:
- Enhance an existing Data Quality Rule
- Merge multiple accepted rules into one data quality rule
- Navigate to the linked data quality rule
Facts about the illustration in the figure:
For many rules from rule mining, an automatic rule implementation in BRFplus is supported.
By adding a usage and preparing it, an active BRFplus rule implementation will be generated.
Information on rules from rule mining is provided in separate sections.
Approving the data quality rule and enabling the usage works as for any other data quality rule.
