Let’s say we want to forecast the energy consumption of a household for August to October 2018.
We have the energy consumption of the household and Average Outdoor Temperature from January 2017 to July 2018. Next, we have the forecasted Average Temperature for August to December 2018 to use as an additional input.
Here is the sample data we’re working with:
|Month / Year||Consumption (kWh)||Average Temperature (Degrees C)|
|August 2018||18.4°C (Forecast)|
|September 2018||14.3°C (Forecast)|
|October 2018||10.1°C (Forecast)|
|November 2018||7.4°C (Forecast)|
|December 2018||2.4°C (Forecast)|
Example 1: Forecast with Additional Variable Input
Forecasting future consumption while making use of the Average Temperature as an Additional Forecast Input produces the following result:
Please note: without the forecasted temperature data for the future period (August to December 2018), we would receive an error message: “None of the additional inputs are being considered in the forecast due to a lack of future values”.
In the above Time Series Chart, the dotted line indicates the Forecast produced by the SAP Analytics Cloud predictive algorithm. The part of the forecast starting in February 2017 is called the Past Period Forecast. The Past Period Forecast helps to indicate the accuracy of the forecast. As does the shaded confidence interval shown above and below the forecasted Consumption for August, September, and October 2018.
The resulting forecasted Consumption data points for August, September, and October 2018 take into consideration the forecasted Average Temperature. Therefore, this forecast is likely more accurate than a forecast based on historical consumption alone.
Example 2: Forecast without Additional Variable Input
SAP Analytics Cloud is also able to generate a forecast without any additional variable input. Forecasting based on historical data alone produces the following result:
In the above Time Series Chart, the dotted line indicates the Forecast produced by the SAP Analytics Cloud predictive algorithm. As the Forecast does not take into consideration the Average Temperature, the forecast is based solely on the historical Consumption value and therefore is likely to be less accurate than in Example 1.
By making use of Additional Forecast Inputs in your Time Series Forecast, you can provide more information to the SAP Analytics Cloud forecasting algorithms, resulting in more accurate predictions.