Manage Consumption Groups

Consumption groups are the connections to a data table in OneStream. Use the Consumption Groups page to manage consumption groups that are used to load model predictions and insights into tables within OneStream.

NOTE: The destination Data Connection will be the same as the specified Data Source Connection for the Target Data Source

Consumption groups can be created after you have specified targets and features and verified your data sets using the Data > Dataset page.

Create Consumption Groups

The consumption groups you create are available when you run a new prediction.

1. Click the Consumption Groups icon in the Sensible Machine learning toolbar. The Consumption Groups dialog box displays.

2. Click Add a New Consumption Group .

3. In the Add Consumption Group dialog box, select the type of consumption to add in the Consumption Type field. This determines the other settings. See the Consumption Group types here.

4. Type a name for your consumption group in the Consumption Group field.

5. Type a name for the Output name in the Output Name field.

6. Make selections from these options (or subset of options based on consumption type). The options you can select include the following values:

   Export Action: The type of action this consumption group should be auto-attached to. If auto-attached, it automatically runs and exports as a part of that type of action. Examples include prediction and pipeline jobs.

   Actuals Types: The type of actuals (engine cleaned, engine uncleaned, source) to include in the consumption group.

   Target Frequency: The frequency to export data in.

   Merge Method: The type of merge to perform on data if there are multiple data points for a date.

   Models to Return: The models to return in the consumption group.

   Prediction Intervals: Indicates if prediction intervals should be included in the export.

   Extraction Type: Select Batch to export only the latest prediction run. Select Time to use Start Date Type and End Date Type fields for the earliest prediction to the latest prediction or custom time frames.

   Start Date Type: If the earliest start date or a custom beginning date should be used.

   Start Date Time, Start Date Hour: The beginning date and hour of the day of the consumption group (inclusive). Available if the If Start Date Type is set to Custom.

   End Date Type: If the latest end date or a custom end date should be used.

   End Date TimeEnd Date Hour: The end date and end hour of the day of the consumption group (inclusive). Available if the If End Date Type is set to Custom.

   Group Name: Type a name for the consumption group to be exported.

   Output Table Name: The name of the outputted table. This is in the same database as the target data set.

   Aggregation Type:.

   

7. Click Save.

Export Consumption Group Data

You can manually export all data defined in a consumption group to the associated new or existing OneStream table. If the table already exists, it must have the correct schema; otherwise, it does not populate the prediction data.

NOTE: Groups can also be automatically exported by setting the Export Action option of a given consumption group.

To export a consumption group:

1. Select a group and then click Export Group

   NOTE: Some consumption types may require you to select what to export data for (all deployed builds or a given build based on build time).

2. Confirm that you want to export the group by clicking Export.

Consumption Group Types

There are multiple types of consumption groups that can be configured to export a variety of information from Sensible Machine Learning. These exports can be used in downstream processes for a variety of applications.

The following list describes the consumption group types and shows the schema for each.

Feature Effect

Description: The feature effect consumption type contains data informing how a given feature and its values compare to a given target's actual values and prediction values. This provides insight into whether a correlation exists between certain feature values and predictions or actuals. See Appendix 4: Interpretability for details.

Schema: Model, FeatureLowerBound, FeatureUpperBound, FeatureAvgValue, PredictionAvgValue, TargetAvgValue, FeatureName, FeatureShortName, TargetName, ConsumptionID, ProjectID, ConsumptionRunID, ConsumptionRunTime, XperimentKernelID, XperimentBuildID, XperimentSetID, BuildInfoID, [TargetDimensions]

Feature Impact

Description: The feature impact consumption type contains data informing how much a given feature influences the model for a given target. A large FeatureImpactValue means that the feature is an important driver for the model predictions for that target. See Appendix 4: Interpretability for details.

Schema: FeatureName, FeatureShortName, FeatureImpactValue, FeatureImpactType, TargetName, ModelName, Category, ConsumptionID, ProjectID, ConsumptionRunID, ConsumptionRunTime, XperimentKernelID, XperimentBuildID, XperimentSetID, BuildInfoID, [TargetDimensions]

Model Forecast Backtest

Description: The model forecast backtest consumption type contains the backtest results (and possibly prediction intervals) made by models for each target from the model build phase of the application.

NOTE: Not all model builds contain a backtest portion. It is dependent on the number of data points.

Schema: Model, ModelCategory, TargetName, Value, LowerPI, UpperPI, Date, ModelRank, ConsumptionID, ProjectID, SplitID, ConsumptionRunID, ConsumptionRunTime, XperimentKernelID, XperimentBuildID, XperimentSetID, BuildInfoID, [TargetDimensions]

Model Forecast Deployed

Description: The model forecast deployed consumption type contains the predictions (and possibly prediction intervals) made by models for each target from the utilization phase of the application.

Schema: Model, ModelCategory, TargetName, Value, LowerPI, UpperPI, Date, ModelRank, PredictionCallID, PredictionScheduledTime, ConsumptionID, ProjectID, ConsumptionRunID, ConsumptionRunTime, ForecastStartDate, ForecastNumber, ForecastName, XperimentKernelID, XperimentBuildID, XperimentSetID, BuildInfoID, [TargetDimensions]

Model Forecast V1

Description: The model forecast v1 consumption type is the same as the model forecast deployed consumption type but keeps the same table schema as Sensible Machine Learning versions SV103 and earlier. This consumption type is deprecated Sensible Machine Learning versions SV200 and later, and should not be used except for maintaining backwards compatibility for existing processes while they are transferred to newer consumption types

Schema: Model, ModelCategory, TargetName, Value, Date, ModelRank, FKPredictionCallID, PredictionScheduledTime, FKConsumptionID, FKProjectID, ConsumptionRunID, ConsumptionRunTime, [TargetDimensions]

Prediction Explanations Backtest

Description: The prediction explanations backtest consumption type contains the amount that the feature influenced (positive or negative) the prediction of a given model for a given target for a given date during the backtest in the model build section of the application. See Appendix 4: Interpretability for details.

Schema: Date, FeatureName, FeatureShortName, PredictionExplanationValue, FeatureValue, PredictionExplanationType, TargetName, Model, ModelStage, ModelCategory, ModelITerationID, ConsumptionID, ProjectID, ConsumptionRunID, ConsumptionRunTime, XperimentKernelID, XperimentBuildID, XperimentSetID, BuildInfoID, [TargetDimensions]

Prediction Explanations Deployed

Description: The prediction explanations deployed consumption type contains the amount that the feature influenced (positive or negative) the prediction of a given model for a given target for a given date during the forecasts in the utilization section of the application. See Appendix 4: Interpretability or details.

Schema: Date, FeatureName, FeatureShortName, PredictionExplanationValue, FeatureValue, PredictionExplanationType, TargetName, Model, ModelStage, ModelCategory, ModelITerationID, ConsumptionID, ProjectID, ConsumptionRunID, ConsumptionRunTime, PredictionCallID, PredictionScheduledTime, ForecastStartDate, ForecastNumber, ForecastName, XperimentKernelID, XperimentBuildID, XperimentSetID, BuildInfoID, [TargetDimensions]