Generating a fraud prediction utilizing a fraud-prediction machine-learning model

A system comprising:

• at least one processor;
• and
• at least one non-transitory computer-readable medium storing instructions that, when executed by the at least one processor, cause the system to:
• access a fraud detection machine-learning model that was trained using one or more training fraud predictions corresponding to training digital claims, wherein training the fraud detection machine-learning model comprises comparing a training digital claim to fraud action labels using a loss function to generate losses that are used to adjust one or more parameters of the fraud detection machine-learning model;
• receive, from a client device associated with a user account, a digital claim indicating that a network transaction associated with the user account was not authorized;
• identify one or more features associated with the digital claim;
• generate, as an output of the fraud detection machine-learning model, a fraud prediction indicating a probability that the digital claim is fraudulent by processing the one or more features associated with the digital claim with the fraud detection machine-learning model;
• provide, for display in a graphical user interface, a visual indicator of the fraud prediction for the digital claim;
• receive a fraud action label indicating whether the digital claim was fraudulent or not fraudulent;
• and
• update the one or more parameters of the fraud detection machine-learning model using the fraud action label by:
• comparing the fraud action label to the fraud prediction and one or more additional fraud action labels to one or more additional fraud predictions;
• determining, based on comparing the fraud action label to the fraud prediction and one or more additional fraud action labels to one or more fraud predictions, that the fraud prediction and the one or more additional fraud predictions do not satisfy a measure of loss;
• and
• updating parameters of the fraud detection machine-learning model.

A method comprising:

• accessing a fraud detection machine-learning model that was trained using one or more training fraud predictions corresponding to training digital claims, wherein training the fraud detection machine-learning model comprises comparing a training digital claim to fraud action labels using a loss function to generate losses that are used to adjust one or more parameters of the fraud detection machine-learning model;
• receiving, from a client device associated with a user account, a digital claim indicating that a network transaction associated with the user account was not authorized;
• identifying one or more features associated with the digital claim;
• generating, as an output of the fraud detection machine-learning model, a fraud prediction indicating a probability that the digital claim is fraudulent by processing the one or more features associated with the digital claim with the fraud detection machine-learning model;
• providing, for display in a graphical user interface, a visual indicator of the fraud prediction for the digital claim;
• receiving a fraud action label indicating whether the digital claim was fraudulent or not fraudulent;
• and
• updating the one or more parameters of the fraud detection machine-learning model using the fraud action label by:
• comparing the fraud action label to the fraud prediction and one or more additional fraud action labels to one or more additional fraud predictions;
• determining, based on comparing the fraud action label to the fraud prediction and one or more additional fraud action labels to one or more fraud predictions, that the fraud prediction and the one or more additional fraud predictions do not satisfy a measure of loss;
• and
• updating parameters of the fraud detection machine-learning model.