Predictive Scoring with Confidence Models
Overview
Running a predictive scoring job with a confidence model is similar to scoring with a predictive training model. The primary difference is in the model URI that is provided when the scoring job is launched. If the URI provided is from a confidence model, the scoring output includes confidence interval information.
Predictive scoring with a confidence model can be run using either a batch service or a real time service. Predictive scoring with confidence models can also be run on both time series and non-time series data. The output of the scoring depends on the type of data being scored. For more information about the output, see
Interpreting Predictive Scoring Results with Confidence Models
Run Scoring and Retrieve Results with Confidence Models
To run scoring with a confidence model, you need a URI for a confidence model. For information about generating a confidence model and retrieving the model URI, see
Generating Confidence Models.
1. In ThingWorx Composer, click the Things icon in the left navigation panel, open the Prediction Thing, and click on the Services tab.
2. Open the BatchScore service.
|
You can also run a RealtimeScore job with a confidence model, but you will not be able to view the results in a batch or save them.
|
3. In the modelUri field, enter results:/models/ and then paste the model ID at the end.
4. Enter the following additional minimum parameters:
◦ jobName – A name to reference the job.
◦ datasetRef – Click to open a dialog box, select Add, and enter the datasetUri and a format (must be parquet or csv).
◦ goalField – Field name of the goal variable.
5. Click Execute. The predictive scoring job runs and outputs a jobId.
6. In the Output panel, copy the jobId.
7. Return to the list of Things, open the Results Thing, and click on the Services tab.
8. Open the SaveResult service and enter the following parameters:
◦ resultType – Enter PREDICTION.
◦ resultId – Paste the copied jobID.
9. Click Execute. The scoring results are saved and the Output panel displays a file path that indicates their location.
10. Copy the filePath, enter it into a new browser tab, and press Enter.
11. When prompted, save the CSV results file.