The xSuite Prediction Server is an AI component based on neural networks. The xSuite Prediction Server can be used to predict data for individual invoices (e.g. account-assignment data and responsible persons) as well as expected totals (number, average amount and total invoice amount per vendor) for upcoming months.

For this purpose, historical invoice data is extracted from the SAP system and used as training set for the prediction models used by the xSuite Prediction Server. The xSuite Prediction Server is an external component that is not part of xSuite Cube Business Solutions and is currently in the testing phase.

In the previous version 5.2.5, the connection to the xSuite Prediction Server was included for the first time and was used for account-assignment proposals and for determining the company code and the vendor of an invoice.

In the current version, this connection has been supplemented with the Invoice Forecast (prediction of the total number of invoices and sum total for invoices from a supplier for future months). This function is available as an xSuite App for SAP Fiori. For more information, see xSuite Invoice Forecast App with the Prediction Server (SAP Fiori).

To train the associated prediction model, historical invoice data for the last 4 years (with reference to a baseline date) must first be submitted to the Prediction Server. The report /WMD/XF_IV_KERAS_FCAST_UPLOAD is used for this purpose:

First, the Prediction Server must be connected via RFC destination. For more information, see the publication xSuite Cube 5.2.5 Release Notes.

Once the model has been calculated, the Prediction Server makes predictions for the upcoming 12 months. These predictions can be retrieved from the SAP system using the /WMD/XF_IV_KERAS_FCAST_GET report, and can be stored in the database.

Then it can be displayed graphically via the xSuite Invoice Forecast App.