Seasonal Linear Regression(SLR) is recently introduced in the IBP Demand 1908 release which calculates the seasonal forecast based on a linear function.
It can take into account trend and seasonality pattern which it identifies in the historical data.
Whereas triple exponential smoothing already in IBP Demand from the initial release that is also used to handle the time series data containing a seasonal component. This method is based on three smoothing equations: stationary component, trend, and seasonal.
Many time it is anticipated that SLR may yield more robust results than the triple exponential smoothing algorithms (which can also be used for trend or seasonality).
Let’s see how both the model perform in determining the forecast on the seasonal data. To compare the predictive power of these model, two seasonal time series sample data were selected.
Created a forecast model for both algorithm using Manage Forecast Models App.
For Graphical Analytical analysis, IBP Analytics Advanced App is used to plot time-series history data and expected forecast data by selecting Create -->Analytics Chart.
Time-Series Data1:
Actual sales Qty, system calculated Ex-post forecast, and generated forecast for 12 periods from both algorithm for time-series data1 plotted as below. Time series plot gives a clear picture of the data-series is having both seasonality and trend component in it.
So here it can be clearly observed that both algorithm do a decent job in the determination of trend and seasonality. Trend and seasonality are very well reflecting in past and future values.
But looking at error indicator MAPE(i.e. mean absolute percentage error) value in business logs indicates that Seasonal Linear Regression algorithm leaves behind triple exponential smoothing algorithm for this time series-data in model fitting which is not so simple to observe through the graphical plot for Ex-post forecast from both algorithms.
Time-Series Data2:
Actual sales Qty, Calculated Ex-post forecast, and generated forecast for 12 periods from both algorithm for time-series data2 plotted as below. Time series plot gives the clear picture that the data series is only having extreme seasonality and with the negligible trend.
So here it can be clearly observed for this time-series data algorithm Triple exponential smoothing does the decent job in comparison to Seasonal Linear Regression.
A graphical plot, as well as MAPE value, is in the business log that gives a clear picture that Triple exponential smoothing does well for this time-series data.
Conclusion
Seasonal Linear Regression is a very powerful algorithm that is recently added in IBP Demand in 1908 release which can also yield more powerful results than the triple exponential smoothing for some of the time-series data as having seen for the time series data 1 in our analysis.
The algorithm perfectly captures both seasonality as well trend to reflect in the future forecast.
References:
https://help.sap.com/viewer/feae3cea3cc549aaa9d9de7d363a83e6/1908/en-US/2a90abfcfc64409ab47b1f8e35c3...
It can take into account trend and seasonality pattern which it identifies in the historical data.
Whereas triple exponential smoothing already in IBP Demand from the initial release that is also used to handle the time series data containing a seasonal component. This method is based on three smoothing equations: stationary component, trend, and seasonal.
Many time it is anticipated that SLR may yield more robust results than the triple exponential smoothing algorithms (which can also be used for trend or seasonality).
Let’s see how both the model perform in determining the forecast on the seasonal data. To compare the predictive power of these model, two seasonal time series sample data were selected.
Created a forecast model for both algorithm using Manage Forecast Models App.
For Graphical Analytical analysis, IBP Analytics Advanced App is used to plot time-series history data and expected forecast data by selecting Create -->Analytics Chart.
Time-Series Data1:
Actual sales Qty, system calculated Ex-post forecast, and generated forecast for 12 periods from both algorithm for time-series data1 plotted as below. Time series plot gives a clear picture of the data-series is having both seasonality and trend component in it.
So here it can be clearly observed that both algorithm do a decent job in the determination of trend and seasonality. Trend and seasonality are very well reflecting in past and future values.
But looking at error indicator MAPE(i.e. mean absolute percentage error) value in business logs indicates that Seasonal Linear Regression algorithm leaves behind triple exponential smoothing algorithm for this time series-data in model fitting which is not so simple to observe through the graphical plot for Ex-post forecast from both algorithms.
Time-Series Data2:
Actual sales Qty, Calculated Ex-post forecast, and generated forecast for 12 periods from both algorithm for time-series data2 plotted as below. Time series plot gives the clear picture that the data series is only having extreme seasonality and with the negligible trend.
So here it can be clearly observed for this time-series data algorithm Triple exponential smoothing does the decent job in comparison to Seasonal Linear Regression.
A graphical plot, as well as MAPE value, is in the business log that gives a clear picture that Triple exponential smoothing does well for this time-series data.
Conclusion
Seasonal Linear Regression is a very powerful algorithm that is recently added in IBP Demand in 1908 release which can also yield more powerful results than the triple exponential smoothing for some of the time-series data as having seen for the time series data 1 in our analysis.
The algorithm perfectly captures both seasonality as well trend to reflect in the future forecast.
References:
https://help.sap.com/viewer/feae3cea3cc549aaa9d9de7d363a83e6/1908/en-US/2a90abfcfc64409ab47b1f8e35c3...
- SAP Managed Tags:
| User | Count |
|---|---|
| 3 | |
| 2 | |
| 2 | |
| 2 | |
| 1 | |
| 1 | |
| 1 | |
| 1 |
