Use an Invoice Object Recommendation (IOR) Dataset Schema to Upload Training Data to Data Attribute Recommendation

You’ll use Swagger UI, via any web browser, to call the Data Attribute Recommendation APIs. Swagger UI allows developers to effortlessly interact and try out every single operation an API exposes for easy consumption. For more information, see Swagger UI.

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. You’ll use all three Swagger UIs throughout the tutorials.

For this tutorial, copy the URL of the Swagger UI for dm and open it in a browser tab.

After finishing this tutorial, keep the Swagger UI for dm open to perform the clean up tasks in Use the Invoice Object Recommendation (IOR) Business Blueprint to Predict Financial Objects.

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.

   Code Copy

   Bearer <access_token>

3. Click Authorize and then click Close.

   

Now, you need to create a new dataset schema. A dataset schema describes the structure of datasets.

Use the following dataset schema:

{
   "features":[
      {
         "label":"BUKRS",
         "type":"CATEGORY"
      },
      {
         "label":"BELNR",
         "type":"CATEGORY"
      },
      {
         "label":"GJAHR",
         "type":"NUMBER"
      },
      {
         "label":"BUZEI",
         "type":"NUMBER"
      },
      {
         "label":"KOART",
         "type":"CATEGORY"
      },
      {
         "label":"WRBTR",
         "type":"NUMBER"
      },
      {
         "label":"LIFNR",
         "type":"CATEGORY"
      },
      {
         "label":"BLART",
         "type":"CATEGORY"
      },
      {
         "label":"BUDAT",
         "type":"CATEGORY"
      },
      {
         "label":"MWSKZ",
         "type":"CATEGORY"
      }
   ],
   "labels":[
      {
         "label":"HKONT",
         "type":"CATEGORY"
      }
   ],
   "businessBlueprintFields":[
      {
         "label":"SHKZG",
         "type":"CATEGORY"
      },
      {
         "label":"BUZID",
         "type":"CATEGORY"
      },
      {
         "label":"BSCHL",
         "type":"CATEGORY"
      },
      {
         "label":"STBLG",
         "type":"CATEGORY"
      }
   ],
   "name":"ior-dataset-schema"
}

The schema is divided into features and labels. The features are the inputs for the machine learning model whereas the labels are the fields that will be predicted. Thus, this schema provides as input training fields such as BUKRS for company code, and wants to predict G/L accounts (HKONT).

To create the dataset schema, proceed as follows:

1. In Swagger UI, expand the endpoint POST /datasetSchemas by clicking on it. Then click Try it out.

   

2. Copy the above dataset schema into the text area. Then click Execute to create it.

   

3. Further below, you find the response of the service. The response includes a representation of dataset schema that was just created. Additionally, the dataset schema received an id. Copy it locally as you’ll need it in the next step.

   

You have successfully created a dataset schema.

Next, you need to create a dataset using the dataset schema that you have created. The dataset is a table that holds the data that you’ll upload later.

To create the dataset, proceed as follows:

1. Expand the endpoint POST /datasets by clicking on it. Then click Try it out.

   

2. In the text area, replace the parameter datasetSchemaId with the id that you copied from the previous step and replace the parameter name with an appropriate name for you dataset, ior_tutorial_dataset, for example. Then click Execute to create the dataset.

   

3. In the response of the service, you find the id of your dataset. Copy it locally as you’ll need it in the next steps and also in the next tutorial: Use the Invoice Object Recommendation (IOR) Business Blueprint to Train a Machine Learning Model. Additionally, you find the status of the dataset. The status is NO_DATA as no data file has been uploaded yet.

   

You have successfully created a dataset.

The final step of this tutorial is to upload data to your dataset.

In this tutorial, you’ll use this dataset. Right click on the link, then click Save link as to open the file dialog. In the dialog, replace the file ending txt with csv as indicated below. Then save the file.

In Swagger UI, proceed as follows to upload to the data:

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

   

2. Fill the parameter id with the id of your dataset that you previously copied.

3. Click Choose File below the parameter Request body. In the dialog that opens, select the IOR dataset that you just downloaded. Then click Execute to upload the data.

In the response, you’ll see that the status of your dataset has changed to VALIDATING. The service is now validating the data that you have uploaded.

You have successfully uploaded data to your dataset.

To check the validation status of your data, proceed as follows:

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

   

2. Fill the parameter id with the id of your dataset. Click Execute.

   

3. In the response of the service, you find the status of your dataset. If the status is still VALIDATING, check back in a few minutes. If the status is SUCCEEDED, your data is valid. In case the status is either INVALID_DATA or VALIDATION_FAILED, create a new dataset and upload the data once again.

   

You have successfully created a dataset and uploaded data. You can now use the dataset to train a machine learning model.