TD_UNNORMALIZE reconstructs a series previously created by TD_SEASONALNORMALIZE. The TD_UNNORMALIZE function is usually used during the forecasting phase of modeling.
Seasonal normalization removes seasonal fluctuations from the data. This involves dividing each observation by the average value of the corresponding season. The transformation enables the detection of other patterns in the data, such as trends or irregular fluctuations, which may not be visible due to the seasonal variations.
After performing seasonal normalization, it may be necessary to un-normalize the data to obtain the original values. Unnormalizing involves reversing the normalization process to obtain the original values of the data.
The following procedure is an example of how to use TD_UNNORMALIZE when the series to be modeled is found to be nonstationary:
- Use TD_SEASONALNORMALIZE to make the series stationary.
- Develop the ARIMA forecast model.
- Use the ARIMA model to forecast the normalized series.
- Use TD_UNNORMALIZE on the forecasted normalized series to undo the effects of normalization and produce the final forecasted series result.