Understanding Machine Learning for Outsorted Billing Documents

Details:

Predictive scenario: UTI_BI_OUTSORTED.

Training and prediction timeframes maintained in table TEML_OUTBILPARAM (transaction SM30).

Months to Sample Start (from Current Date): the months to sample start from current date represent the relative period in months that is used by the Utilities Machine Learning use cases for the training and prediction set. Example: the current date is 05/05/2019. If the value entered in the field Months to Sample Start is 12 (months), the Machine Learning algorithm uses the date 05/05/2018 as the start date for the training and prediction set.

Months to Sample End (from Current Date): the months to sample end from current date represent the relative period in months that is used by the Utilities Machine Learning use cases for the training set.

Example: the current date is 05/05/2019. If the value entered in the field Months to Sample End is 2 (months), the Machine Learning algorithm uses the date 03/05/2019 as the end date for the training.

Predictive results integrated with the billing process via report, consumed through ISLM.

The model UTI_BI_OUTSORTED works as follows.

From a process view, the machine learning solution consists of 4 major building blocks:

1. Provision of training and apply-relevant datasets to access and utilize machine learning-relevant information for outsorted billing documents. The features used can be grouped as follows:

   • Segmentation features (master data such as portion, billing class)

   • Amount-based features (for example billing amount, previous billing amount)

   • Statistical features (for example the average billing amount of the last 2 years)

2. Apps Intelligent Scenario Management (App ID F4470) and Intelligent Scenarios (App ID F4469) to access and train the scenario/ model UTI_BI_OUTSORTED.

   Documentation is available here.

3. The report REML_BILLING_OUTSRTD_RELPRED to determine the release confidence values for outsorted billing documents. Automatic release of outsorted billing documents, based on the release confidence against a given threshold.
4. Display the release confidence in the Fiori app Resolve Outsorted Billing Documents (F2186).

The complete documentation about Machine Learning: Process Outsorted Billing Documents | SAP Help Portal is available in the SAP Help Portal.

• 3241037 – IS-U Composite SAP Note for Machine Learning

• 3345830 – Machine Learning for IS-U Billing

• 3480987 – Performance in IS-U ML for Outsorted Billing Documents: Improved runtime and reduced memory usage. Dependency on APL version applies for S/4HANA 2023 or lower.

• C_OutsrtdUtilsBillgMaLeTrng – Training dataset

• C_OutsrtdUtilsBillgMaLeApply – Apply dataset

• Faster resolution or reversal of outsorted billing documents.

• Increased billing stability and accuracy.

• Enhanced cash flow through reduced delays in invoice processing.

• Creation, training, and activation of a model version in ISLM.

• Correct setup of ISLM (SAP Help – ISLM Initial Setup).

• SAP Note 2631182 Central Release Note Predictive Analytics integrator (PAi) / Intelligent Scenario Lifecycle Management (ISLM)

SAP Note 3480987 introduces a revised Machine Learning model for outsorted billing documents: UTI_BI_OUTSORTD_DOC.

This model replaces the earlier UTI_BI_OUTSORTED scenario and provides significant performance improvements in terms of memory usage and runtime efficiency.