PredictorDataTables Table

This table serves as an index to the source data tables and an attributes table. For each model configuration, there must be at least one Dense (Type 0) or Sparse (Type 1) source data table, and one Attributes (Type 2) table. Only one Dense table is allowed, but there can be multiple Sparse tables. A blank attributes table is created if one is not specified.

Column Name Type Description Required?
ModelCfgName DBTYPE_WSTR Model configuration name (foreign key to the PredictorModelCfgs table). Yes
TableName DBTYPE_WSTR Table name. Yes
Type DBTYPE_UI2 Type of table:
  • 0 = Dense (1 row per user, such as for demographic information)

  • 1 = Sparse (1 row per transaction, such as for product purchases)

  • 2 = Attributes
Yes
AggregateColumn DBTYPE_WSTR Column on which the PivotColumn is aggregated.

For example, QTY if the QTY column represents the quantity of each SKU purchased.

Yes, if Type = Sparse.

Otherwise should be NULL.

AggregateOperation DBTYPE_UI2 Valid values are:
  • 0 = SUM

  • 1 = MAX

  • 2 = MIN

  • 3 = AVG

  • 4 = COUNT
Yes, if Type = Sparse.

Otherwise should be NULL.

CaseColumn DBTYPE_WSTR Column in TableName where the case ID (usually UserID) is stored. Yes, if Type = Dense or Sparse.

Not applicable if Type = Attributes.

PivotColumn DBTYPE_WSTR Column in TableName that specifies the transactional property to pivot.

For example, SKU if the SKU column contains the products a user purchased.

Yes, if Type = Sparse.

Otherwise should be NULL.


All rights reserved.