Feedback Loops & Label API

At its heart, machine learning algorithms learn from good and bad examples. By default, our machine learning solutions learn from transaction states but you can finetune their performance using labels. Create positive and negative transaction labels to give more accurate feedback and rule out false positives.

By default, SEON's machine learning solutions learn from transaction states. As a result, both solutions are trained and re-trained using transaction data unique to your account to prevent fraud specific to your business.

Labels

Consider labels as sub-categories of the three main states defined in SEON by default. You can use labels to improve the granularity of the information you provide to the machine learning algorithms. 

Positive and negative labels will group similar transactions together and help the AI find connections and correlations between transactions with the same labels.

Tip: You can create new positive or negative labels on the settings page. All transactions can be labeled with previously created labels. 

Manual labeling

Your team can easily add labels to transactions on the Transaction list page. Click on the state of any transaction and select the desired label from the dropdown menu.

Tip: You can also add labels to similar transactions using multi-select. Click the checkbox next to each transaction you want to edit and use the options at the bottom of the screen to set the desired label.

Label API

The Label API offers a completely automated option for providing your machine learning algorithms with feedback. You can also use it as a way of sharing information that would otherwise be unavailable to the algorithms, e.g. the postback data from the payment gateway: is a card authorized, lost, or stolen, etc.

The automation offered by the Label API is especially powerful if you handle hundreds or thousands of transactions with SEON each day.

Make sure to always catagorise labels that are sent in via Label API into one of the categories under the positive or negative label groups of the Machine Learning Settings. Please note that machine learning features such as the Blackbox score or Network scores are only trained on transactions that have been marked with labels added as positive or negative in the settings.

You will need help from your developers to get the Label API up and running. Integration can be done in a matter of hours by seasoned professionals. Check the API reference for the details.

The Machine Learning tab of the Settings page houses all settings related to your whitebox and blackbox models, including labels.

To create a new label simply type its name into the Negative Labels or Positive Labels field and hit enter or click the plus sign. The label will now appear in the drop-down on the Transactions list and be used in the machine learning model training.

Click the toggle switch below the labels setting if you'd prefer SEON's whitebox machine learning solutions train only using the Labels you set up.

Providing valid feedback labels to transactions will enable you to harness the power, speed, and accuracy of SEON machine learning. Marking transactions verified as negative (e.g., fraudulent, thus legitimate to decline) or positive (as legitimate to approve) as quickly as possible enables SEON's machine learning capabilities to learn the exact patterns your business experiences. Labels unlock Blackbox Score fraud prediction models, ADF Network Score models, and Machine Learning Rules to be customer-specific and to perform better.

Learn more: Overview: Read about whitebox machine learning in SEON.
Overview: Learn how blackbox machine learning works.
Guide: Learn how to turn on blackbox scoring for your account.