For DDM, the original data is usually transformed into a data format suitable for further processing. In addition, additional input data (input features) can be constructed automatically, partly based on the original data. This is achieved using Input Feature Construction, which incorporates various transformations and calculations to create additional input features that are suited for use in the model building process. As a result of this procedure, input representations are obtained that are suitable for the particular forecasting task at hand, and which contain valuable information that can be used by the forecasting system.

Obtaining the right data and representing it in the right format is essential in achieving high forecasting performance. Use AdaptiCast Input Feature Construction to automatically provide tailored special-purpose input variables for your forecasting application.