- SAP Managed Tags:
Introduction:
Machine Learning is a term used quite frequently for almost every SAP product. So how come SAP Profitability and performance management (PaPM) product leave it. SAP PaPM, has started with Forecasting function, but now it has four more machine learning functions i.e. Time Series forecast, Classification, Regression, Recommendation and Clustering. It basically leverages the HANA PAL libraries for ML algorithms to perform specific functions.
Today I am presenting Time Series forecasting machine learning function. In any machine learning function data plays an important part. So it should be advisable that we should have good historical data(Around 2-3 years) to have most accurate forecast.
Now consider a scenario for Product profitability forecasting. Business would be interested to know, how are different product lines are profitable. However they even more interested to understand certain insights like how is the forecasted profitability. So that they can take sizable actions.
Let’s understand this through SAP PaPM time series forecasting function.
System Landscape: SAP PaPM
Configuration: We will start the configuration form the last output for profitability results. That becomes the input for Time Series Forecasting Machine Learning function. Below Model table has profitability by product. It also has posting date as one of the date field. Date field is an important parameter for time series function configuration.
Now Machine Learning Function provided with this input.
In the Rule tab, rule type set a “Forecast”
For Date field it is important that it has “DATS” data type. Otherwise system will show date field missing error message. Signal field is an important parameter and it becomes the source field for forecasting.
Exclude the fields which do not have any impact on forecasting result. Provide Forecast granularity as “Year”, “Quarter” or “Month”. I have kept it “Month” in this example. Similarly define how many periods should be forecasted. I kept 3 here.
Similarly segmented by should be define the characteristics relevant for forecasting.
Finally Forecast field, where forecasted results will be posted. This can be same signal field or separate field.
Now activate and run the function.
Now here as per the configuration of end date as “20220221”. System has forecasted the profitability for next 3 period.
Conclusion:
Machine learning function is a great feature to have insights that helps business to act upon specific items. However, it should be integrated with ML operations, so that models can be trained with latest data and provide most accurate prediction.
References:
https://help.sap.com/doc/19bc88aaa06f4f23b582aec2e991bd27/3.14/en-US/loiob989bd56bff66a15e10000000a4...
Thank you.
Machine Learning is a term used quite frequently for almost every SAP product. So how come SAP Profitability and performance management (PaPM) product leave it. SAP PaPM, has started with Forecasting function, but now it has four more machine learning functions i.e. Time Series forecast, Classification, Regression, Recommendation and Clustering. It basically leverages the HANA PAL libraries for ML algorithms to perform specific functions.
Today I am presenting Time Series forecasting machine learning function. In any machine learning function data plays an important part. So it should be advisable that we should have good historical data(Around 2-3 years) to have most accurate forecast.
Now consider a scenario for Product profitability forecasting. Business would be interested to know, how are different product lines are profitable. However they even more interested to understand certain insights like how is the forecasted profitability. So that they can take sizable actions.
Let’s understand this through SAP PaPM time series forecasting function.
System Landscape: SAP PaPM
Configuration: We will start the configuration form the last output for profitability results. That becomes the input for Time Series Forecasting Machine Learning function. Below Model table has profitability by product. It also has posting date as one of the date field. Date field is an important parameter for time series function configuration.
Now Machine Learning Function provided with this input.
In the Rule tab, rule type set a “Forecast”
For Date field it is important that it has “DATS” data type. Otherwise system will show date field missing error message. Signal field is an important parameter and it becomes the source field for forecasting.
Exclude the fields which do not have any impact on forecasting result. Provide Forecast granularity as “Year”, “Quarter” or “Month”. I have kept it “Month” in this example. Similarly define how many periods should be forecasted. I kept 3 here.
Similarly segmented by should be define the characteristics relevant for forecasting.
Finally Forecast field, where forecasted results will be posted. This can be same signal field or separate field.
Now activate and run the function.
Now here as per the configuration of end date as “20220221”. System has forecasted the profitability for next 3 period.
Conclusion:
Machine learning function is a great feature to have insights that helps business to act upon specific items. However, it should be integrated with ML operations, so that models can be trained with latest data and provide most accurate prediction.
References:
https://help.sap.com/doc/19bc88aaa06f4f23b582aec2e991bd27/3.14/en-US/loiob989bd56bff66a15e10000000a4...
Thank you.
