TD_SHAP function has two input tables: InputTabe and ModelTable.
| Table | Description |
|---|---|
| InputTable | The table that contains sample data for which to predict feature contribution. |
| ModelTable | Table with same schema as OutputTable of TD_DecisionForest / TD_XGBoost function for trees or with same schema as OutputTable of TD_GLM function for linear regression. |
InputTable Schema
| Column | Data Type | Description |
|---|---|---|
| accumulate_columns | Any | Column appears once for each specified accumulate_column. Column to copy to output table. |
| input_column | INTEGER, BIGINT, SMALLINT, BYTEINT, FLOAT, DECIMAL, NUMBER | Column appears once for each specified input_column. Predictor variable. If the Detailed argument is set to true, two more columns are appended to the output schema. This restricts the number of columns supported by maximum number of input columns (2048) supported by Teradata - 2 (2046). |
| id | Any | Identifier for each row of input data. |
ModelTable Schema
The model table schema is the same as OutputTable for the following training functions:
- TD_GLM. See TD_GLM Output.
- TD_DecisionForest. See TD_DecisionForest Output.
- TD_XGBoost. See TD_XGBoost Output.