Attributes

Choose the attributes that you want to use for a data set to influence spend classification. To incorporate the unique aspects of your business and process flows, use existing attributes or add up to 20 new extended attributes in a data set.

Once you set up the attributes, those attributes are included in the transaction details and in the batch after classification. You can also exclude any of the standard data fields used during classification if these attributes interfere with your classification results. After you change attribute mapping, create your knowledge bases again to incorporate what you learned from the new attributes into the knowledge base.

Considerations for Defining New Attributes

Consider these factors when you define and include, or exclude attributes:

• Recreate the knowledge base after you change the attribute mapping. A knowledge base created before the change in mapping won’t include the learning from the additional classification attributes.
• Use existing training data to help tag the correct categories to the sample training set.
• Disable a standard attribute with caution as it may lead to unexpected classification results.
• When you enable or disable standard attributes, it doesn’t affect the existing knowledge base. Improve or rebuild the knowledge base for the changes to take place.

Defining New Attributes

The Attributes tab contains the list of all the standard attributes that are extracted for different data sets. Along with these standard attributes, 20 additional attributes per data set are also available to you for influencing classification results.

Here’s how you include additional attributes to use in a data set, that will influence spend classification:

1. In the attribute row, map the source table and source column that you want to be extracted for transactions within a data set.
2. To use this attribute for classification, select Yes in the Use for Classification drop-down list.
3. Prepare the training set, including the additional attributes:
   1. Create a sample training set to extract values for these new attributes for the data set. Note that you should use a smaller sample volume percentage when a data set is significant in size.
   2. Download the sample training set.
   3. Manually assign the correct category to each spend record in one of the auto code columns, based on the taxonomy you’re using.
   4. Save the data set with a different name.
   5. Upload the data set.
4. Create the knowledge base using this training data set. Now the defined attributes are also used as input criteria to train the knowledge base.
5. Start classifying the data.
6. Review results in the batches and correct manually if required.