Before creating an application, consider the purpose of the app and review the template options—Predictor, What-if, or Optimizer. Templates only determine the initial configuration of the application and selecting a template does not mean the app can only be used for that purpose.
The table below describes each template option:
|Predictor||Makes predictions for a target feature based on the information provided when the app is created and deployed.||Hides the What-if and Optimizer widget.|
|What-if||Creates and compares multiple prediction scenarios side-by-side to determine the option with the best outcome.||Displays the What-if and Optimizer widget with only the what-if functionality enabled.|
|Optimizer||Runs simulations to optimize an outcome for a given goal. This is most effective when you want to optimize for a single row.||Displays the What-if and Optimizer widget with only the optimizer functionality enabled.
The All Rows widget displays an Optimized Prediction column.
Create an application¶
To create an application:
- Navigate to the Applications tab.
The available application templates are listed at the top of the page. Click Use template next to the template best suited for your use case.
A dialog box appears, prompting you to name the application, enter a brief description, and choose an access option—Invited Users Only automatically generates a link that can be shared with non-DataRobot users and Anyone outside your network limits sharing to other users, groups, and organizations. The access option determines the initial configuration of the sharing permissions, which can be changed in the application settings.
Click Next: select data source.
Select a deployment for the application and click Create. Note that you must be an owner of the deployment in order to launch an application from it.
After signing in with DataRobot and authorizing access, you are taken to the Applications tab while the application builds.
Create an application from a deployment¶
Applications can also be launched from an existing model deployment.
Navigate to the Deployments inventory and select the deployment you want to launch the application from.
Select Create Application from the action menu of your desired deployment.
To generate a link to share an application with non-DataRobot users, select Anyone outside your network in the Allow access for dropdown menu.
Select the application template you would like to use and click Next: add app info.
Name the application, enter a brief description, and choose an access option. When you're done, click Create.
The application is available for use on the Applications tab.
Deployments with an association ID¶
When creating an application from a deployment with an association ID, note the following:
- Accuracy and data drift are tracked for all single and batch predictions made using the application.
- Accuracy and data drift are not tracked for synthetic predictions (simulations) made in the application using the What-If and Optimizer widget.
- You cannot add an association ID to deployments that have already been used to create an application.
In the deployment Settings, add an association ID. If Require association ID in prediction requests is enabled, this setting cannot be disabled after the application is created.
If an application is created from a deployment with an association ID, the association ID is added as a required field to make single predictions in the application. This field cannot be removed in Build mode.
Customize apps with an association ID
With your application open, click Build and select Prediction Details (2) from the Editing page dropdown.
Select the Row Identifier widget at the top of the page. Select the association ID from the dropdown on the left to display this value on the prediction results page for each prediction.
Time series applications¶
You can create Predictor applications from time series deployments—single series and multiseries—allowing you to view the following insights in the Time Series Forecasting widget:
- Predicted vs Actual: Displays a prediction, actual (if uploaded), and forecast line, as well as a bar graph representing residuals (the difference between the prediction and actual).
- Prediction Explanations: Displays Prediction Explanations over time for each prediction.
Note the following before creating a time series application:
- The project uses
yearsas the time unit.
- To include calendar events in the widget, add a calendar file to your project and include calendar events for the timeline of the training dataset and forecasting window.
- Known in advance (KA) features must be set during project creation.
- You cannot train a model on the Time Series Informative Features list after it has been deployed to production.
- The project must be deployed to a DataRobot prediction server.
- The deployment must have an association ID and the appropriate deployment settings configured.
Configure a time series deployment¶
When creating an application from a time series deployment, there some additional settings required. To configure the time series deployment, go to the Deployment inventory, select a time series deployment in the deployment, and navigate to the Settings tab.
Use the table below to configure the appropriate deployment settings for a time series Predictor application:
|Association ID||Required. Enter the Series ID in the Association ID field.|
|Require association ID in prediction requests||Toggle on to require an association ID in batch predictions. This prevents you from uploading a dataset without an association ID, which may affect the accuracy of the predictions.|
|Enable target monitoring||Required. Must be toggled on for time series applications.|
|Enable feature drift tracking||Required. Must be toggled on for time series applications.|
|Enable automatic actuals feedback for time series models||Toggle on to have DataRobot add actuals based on prediction file data.|
|Track attributes for segmented analysis of training data and predictions||Required. Must be toggled on for time series applications.|