Operation error detection

A method comprising:

• acquiring operation data associated with a plurality of invoice submissions;
• for each of the plurality of invoice submissions, generating a set of feature values based on associated operation data;
• determining a first set of the plurality of invoice submissions associated with an error;
• for each operation of the first set of the plurality of invoice submissions:
• automatically determining a rejection reason based on the operation data associated with the invoice submissions, the operation data comprising error text indicating why a particular invoice submission of the plurality of invoice submissions was rejected;
• parsing the error text to identify one or more pre-defined string patterns in the error text and determining a rejection label associated with the rejection reason based on the identification of the one or more pre-defined string patterns;
• associating the rejection label with the set of feature values generated for the invoice submissions;
• training a model to output a label based on the sets of feature values generated for each operation of the first set of the plurality of operations and associated rejection labels;
• receiving invoice data;
• generating a first set of feature values based on the invoice data;
• inputting the first set of feature values to the trained model;
• receiving a label from the trained model;
• and
• presenting a message associated with the received label to a user on a display screen, the message indicating one or more reasons for rejecting an invoice comprising the invoice data.

A non-transitory medium storing program code executable by at least one processing unit of a computing system to cause the computing system to:

• acquire operation data associated with a plurality of operations;
• for each of the plurality of operations, generate a set of feature values based on associated operation data;
• for each of the plurality of operations:
• automatically determine a rejection reason based on the operation data associated with the operation, the operation data comprising error text indicating why a particular distinct document was rejected;
• parse the error text to identify one or more pre-defined string patterns in the error text and determine a rejection label associated with the rejection reason based on the identification of the one or more pre-defined string patterns;
• associate the rejection label with the set of feature values generated for the operation;
• train a model to output a label based on the sets of feature values generated for each operation of the plurality of operations and associated rejection labels;
• receive invoice data;
• generate a first set of feature values based on the invoice data;
• input the first set of feature values to the trained model;
• receive a label from the trained model;
• and
• present a message associated with the received label to a user on a display screen, the message indicating one or more reasons for rejecting an invoice comprising the invoice data.