Use the Regression Model Template to Predict Data Records

In the service key you created for Data Attribute Recommendation in the previous tutorial: Use Free Tier to Set Up Account for Data Attribute Recommendation and Get Service Key or Use Trial to Set Up Account for Data Attribute Recommendation and Get Service Key, you find a section called swagger (as highlighted in the image below) with three entries, called dm (data manager), mm (model manager) and inference.

For the following step, copy the URL of the Swagger UI for inference and open it in a browser tab. The Swagger UI for inference allows you to predict data records using your machine learning model that you have created in the previous tutorial: Use the Regression Model Template to Train a Machine Learning Model.

1. To be able to use the Swagger UI endpoints, you need to authorize yourself. In the top right corner, click Authorize.

   

2. Get the access_token value created in the previous tutorial: Get OAuth Access Token for Data Attribute Recommendation Using Any Web Browser, then add Bearer (with capitalized “B”) in front of it, and enter in the Value field.

   Bearer <access_token>
   

3. Click Authorize and then click Close.

   

To predict data records, proceed as follows:

1. Expand the endpoint POST /models/{modelName}/versions/1 by clicking on it. Then click Try it out.

   

2. In the parameter modelName, enter your model name, regression_tutorial_model, for example.

3. In the parameter body, you have to provide the data you want to predict. According to the dataset schema that you have created in Use the Regression Model Template to Upload Data to Data Attribute Recommendation with Swagger UI, the machine learning model takes the manufacturer and description of the product as input and predicts the price of the product. Replace the text in the parameter body with the following:

   JSON

   Copy

   {
      "topN":1,
      "objects":[
         {
            "objectId":"optional-identifier-1",
            "features":[
               {"name":"manufacturer", "value":"Energizer"},
               {"name":"description", "value":"Alkaline batteries; 1.5V"}                
            ]
         },
         {
            "objectId":"optional-identifier-2",
            "features":[
               {"name":"manufacturer","value":"Eidos"},
               {
                  "name":"description",
                  "value":"Unravel a grim conspiracy at the brink of Revolution"
               }                
            ]
         }
      ]
   }
   

4. Click Execute to send the above input for prediction to the service.

   

In the response of the service, you find the values that the model predicted.

The regression model template produces a different output when compared with the other Data Attribute Recommendation model templates and business blueprints. Instead of providing a pair value/probability result, it returns a pair value/std.

The regression model template is not deterministic. The value is the average of several hundred predictions made by the model for the same input and std is the standard deviation of these predictions. The std should give a feeling of how certain the model is about the predicted value. The smaller the std, the more confident the model prediction.

For example, in the first result below (“value”: “233.9787445068”), the model has predicted the average price of 234 dollars. Most of the model predictions lie in the area between 201 (234-33) and 267 (234+33) dollars. Therefore the model is quite confident and even more accurate in the second result (“value”: “137.1488037109”) that has a standard deviation of only 19.

{
  "id": "9f989bcf-2275-4682-654a-d5ea1acba0a9",
  "predictions": [
    {
      "labels": [
        {
          "name": "price",
          "results": [
            {
              "std": 33.0863685608,
              "value": "233.9787445068"
            }
          ]
        }
      ],
      "objectId": "optional-identifier-1"
    },
    {
      "labels": [
        {
          "name": "price",
          "results": [
            {
              "std": 18.7408504486,
              "value": "137.1488037109"
            }
          ]
        }
      ],
      "objectId": "optional-identifier-2"
    }
  ],
  "processedTime": "2022-05-10T13:00:23.024141",
  "status": "DONE"
}

You have successfully predicted the price of product data records. Feel free to adapt the examples above and retry the prediction.

Now that you have learned the whole process about how to use the Regression Model Template from the Data Attribute Recommendation service, it’s time to clean up. This way, the technical limits won’t get in your way when trying out other Data Attribute Recommendation tutorials. See Technical Constraints and Free Tier Option Technical Constraints.

First, undeploy your model. For that, go back to the Swagger UI for mm and:

1. Expand the endpoint DELETE /deployments/{deploymentId} by clicking on it. Then click Try it out.

   

2. Fill the parameter deploymentId with the ID of your deployment. Use the GET /deployments endpoint in case you no longer have the deployment id in hand.

   

If the response code is 204, the model has been successfully undeployed.

You have successfully undeployed your model, but the model is not yet deleted. Instead it isn’t in production which means that you cannot make inference requests. You can deploy it again at any time using the POST /deployments endpoint.

Once undeployed, you can delete your model.

1. Expand the endpoint DELETE /models/{modelName} by clicking on it. Then click Try it out.

   

2. Fill the parameter modelName with the name of your machine learning model (regression_tutorial_model). Use the GET /models endpoint in case you no longer have the model name in hand.

   

If the response code is 204, the model has been successfully deleted.

Now that the model is deleted, you can delete the training job that created the model.

1. Expand the endpoint DELETE /jobs/{jobId} by clicking on it. Then click Try it out.

   

2. Fill the parameter jobId with the ID of your training job. Use the GET /jobs endpoint in case you no longer have the job id in hand.

   

If the response code is 204, the training job has been successfully deleted.

To clear the uploaded data, you can now delete the dataset as the associated training job is already deleted. For that, go back to the Swagger UI for dm and:

1. Expand the endpoint DELETE /datasets/{id} by clicking on it. Then click Try it out.

   

2. Fill the parameter id with the ID of your dataset. Use the GET /datasets endpoint in case you no longer have the dataset id in hand.

   

If the response code is 204, the dataset has been successfully deleted.

If you do not need your dataset schema anymore, you can delete it as well.

1. Expand the endpoint DELETE /datasetSchemas/{id} by clicking on it. Then click Try it out.

   

2. Fill the parameter id with the ID of your dataset schema. Use the GET /datasetSchemas endpoint in case you no longer have the dataset schema id in hand.

   

If the response code is 204, the dataset schema has been successfully deleted.

Congratulations, you have completed this tutorial.