This last article on defining data is about some special columns that needs to be in your model.

There are 5 different types of special columns, that are treated in a special and predefined way by nbt.

  • Id. This is the identifier of the row. It needs to be there, so the score of every record can be properly set. The output of the analysis is a score for each Id, hence we need to be able to identify them somehow. This value must be unique, at least, in all the records that NBT should calculate a scoring.
  • Contact Date (optional) : This relates to the date of the information contained in each row. If this column is not defined, there will not be a Weekly Success Chart in the reports section.
  • Target : column that defines the objective of the predictive. It should binary. 1 means positive row, 0 means negative row
  • Valid (optional): column that defines if the row is considered valid and will be used to calculated the predictive models or not. 1 means valid row and 0 means not valid row. Examples of not valid rows are calls that are not answered or people not found in contact center campaigns. If no valid column is set, NBT consider valid every record for calculating predictive models.
  • toScore (optional): column that defines which rows will be scored and which rows will be used as training data for predictive models. Usually, the historical data is used for training and the actual (or future) data is used to calculate scoring. 1 means row that should be scored, it means, the user wants to know a prediction. 0 means that row should not be scored because the target is already know, this row is used to train the predictive models (getting the kwoledge of the historical information).