Inference Process Preparation
Before the inference process of the Cash Application can start, the normal FI-CA clearing process takes place. First, the Bank Statement Items need to get imported and the Payment Lots have to be created – this happens classically with the Transactions FF_5 and FPB7.
The Payment Lot Items can be displayed with Transaction FP05 – as we can see in the slide. For all the items of the payment lot which cannot be cleared against an open item automatically with the existing clearing rules, a clarification case gets created and the inference process with SAP Cash Application starts.
Automatic Clearing
When the Automatic clearing of incoming payment with Machine Learning is switched on for the company code, changing clarification cases in Transaction FPCPL is not possible right away as the Clarification Case first of all gets reserved for the Machine Learning process and therefore manual clarification is not yet allowed.
Cash Application: Transfer Clarification Cases
Cash Application: Transfer Clarification Cases - Transaction FPML_CASHAPP_INF
A scheduled job (inference job) retrieves open clarification cases for payment lot items that could not be cleared automatically and pushes them into the prediction API. When the data volume of the inference data is high, several CSV files are sent to the Machine Learning server.
Cash Application: Process Returned and Reserved Clarification Cases
Cash Application: Process Returned and Reserved Clarification Cases - Transaction FPML_CASHAPP_PROP
Using this program, you can process further the jobs that were generated (inference jobs) with the Transfer Clarification Cases job. You reference the Transfer Clarification Cases by the respective date and run ID.
Several CSV files (with various job IDs) are sent to the Machine Learning server for each run date and run ID when there are large data volumes. A status is designated for each job ID.
If you want to request the proposals from the Machine Learning server, you have to select data (jobs) sent successfully. You have the following options:
| Cash Application: Adjust Confidence Rating |
|---|
| You want to check uncertain proposals automatically for correctness. In the event of success, the confidence rating of these proposals is to be raised to a value that leads to display in manual clarification processing. |
| Additionally, you want to check very certain proposals automatically for correctness. In the event of errors, the confidence rating of these proposals is to be lowered to a value that prevents automatic posting, but it leads to display in manual clarification processing. |
| Cash Application: Clear Automatically |
|---|
| If the proposal achieves at least the confidence rating for clearing (very certain proposals), an automatic posting is triggered. If the outcome is successful, manual clarification case processing is no longer necessary. |
| Cash Application: Release Clarification Cases |
|---|
| You request the proposals and release the clarification cases if the proposals cannot be requested. By releasing the clarification cases, you can enable manual processing, as is usual without using Machine Learning. |
If you select data (jobs) transferred with errors, you have two options:
| Release Clarification Cases |
|---|
| You suppose that a clarification case transferred again will still contain errors. When the clarification cases are released, this allows for the processing to be done manually as this is usually done using Machine Learning. |
| Delete Jobs |
|---|
| You start a new transfer of the clarification case with the Transfer Clarification Cases job. Afterwards, you can delete the jobs containing errors for the former mass run. |
In the Cloud you have to use the Application Jobs for all of the jobs that have been shown before. Those Application Jobs do come with Templates which work very similar to what has been shown for the On-Premise World.
Log Display: Cash Application: Process Returned and Reserved Clarification Cases
This slide shows the Log of the Cash Application: Process Returned and Reserved Clarification Cases Job.
Clarifying Incoming Payments with ML Proposals
For all clarification cases where a Machine Learning Proposal has been provided but no automatic posting has taken place, you can see the proposal with a prediction rate in the Transaction FPCPL – Clarify Incoming Payments. From there you can decide if the proposal looks good and should be posted, or if a different clearing should take place.
