Data cleanup recommendations in preparation for Intelligent Account Combination Defaulting

Summary: We are preparing to implement Intelligent Account Combination Defaulting however our historical set of invoice data contains a lot of errors and we don't want the model to train on incorrect data. We are planning to have accounting resources review coding over the next three months to improve data quality but want to know what other Oracle customers have done to prepare to use this feature effectively. What actions did you take to clean up your historical data (ex. cancelling invoices an re-creating them with correct coding, etc)? How did you approach this work? Was it successful?

Content (please ensure you mask any confidential information):