The goal of this processor is the assessment of different forecast methods and the selection of the best forecasts per forecast step and group.
The processor requires two input tables. The right side input table contains a column with the names of the different forecast methods. The left side input table contains the true values to compare forecasts to, a time key index column, a start index column, a forecast step column and a column for each forecast method with different forecast values.
Warning: The columns referring to the forecast methods in the first input table should match the methods the second input table (same number and same method name).
The Processor delivers two output tables:
- Selected Forecasts: returns the selected forecasts along with the corresponding key, step and the calculated values for each chosen method (winning method, mean method or weighted mean method).
- Metrics: returns a more detailed table with the different computed residuals and weights for each forecast method selected.
Warning: For the second output table to be generated. The "Compute Residuals and Weights" button should be toggled in the configuration.
In this example, we use random, manually entered values. For this purpose, five forecast metrics are used.