When it is used as partition-by-key function to train each partition as a single model, the function accepts the following inputs:
- InputTable is required, containing an input dataset to be used for training GLM per segment model.
- AttributeTable is optional, defining a subset of features to be used with respect to each partition.
- ParameterTable is optional, defining a subset of parameters to be used with respect to each partition.
The ORDER BY clause can be optionally applied to the InputTable to guarantee the result in each run is deterministic. The situation of indeterministic result in a partition can occur if the BatchSize argument is less than the number of rows in the partition. Adding the ORDER BY clause ensures the result is deterministic regardless of the BatchSize settings. The ORDER BY clause can affect performance.