Required Feature Group Types
To create a model monitor, you will need to upload datasets of the following type(s):
| Feature Group Type | Required | Description |
|---|---|---|
| Training Data Table | True | Dataset representing the distribution of features encountered by a model during training. |
| Prediction Log Table | True | Dataset representing the distribution of features encountered by a model during prediction or deployment. |
Feature Group: Training Data Table
Dataset representing the distribution of features encountered by a model during training.
| Feature Mapping | Feature Type | Required | Description |
|---|---|---|---|
| TARGET | Y | The target value the model is training to predict. | |
| MODEL_VERSION | categorical | N | The unique identifier of the model version that was trained with this training row. |
Feature Group: Prediction Log Table
Dataset representing the distribution of features encountered by a model during prediction or deployment.
| Feature Mapping | Feature Type | Required | Description |
|---|---|---|---|
| PREDICTION_TIME | timestamp | N | Timestamp of the prediction. |
| ACTUAL | N | The ground truth value of this prediction data. | |
| PREDICTED_VALUE | Y | Model output value for this prediction data. | |
| PREDICTED_PROBABILITY | N | Model output probability value for prediction data. | |
| MODEL_VERSION | categorical | N | The unique identifier of the model version corresponding to the prediction. |