Abstract: Disclosed are systems and methods for predicting tariff classifications. Item descriptions of historical items and associated tariff codes/classifications are submitted to a machine learning platform, thereby creating a trained model for predicting tariff classifications. The trained model receives product data associated with new items and is utilized to generate predicted tariff classifications, which correspond to particular tariff rates. In examples, a confidence/accuracy score is applied to the prediction. The predicted tariff codes are then analyzed in view of a set of safeguard rules. If any of the predictions fail the safeguard rules, a manual user review may be triggered. Manually corrected feedback is provided for use in subsequent retraining so that the machine learning model accuracy is improved over time. If any of the predictions pass the evaluation under the safeguard rules, the predictions may be automatically approved.

Abstract: Disclosed are systems and methods for predicting tariff classifications. Item descriptions of historical items and associated tariff codes/classifications are submitted to a machine learning platform, thereby creating a trained model for predicting tariff classifications. The trained model receives product data associated with new items and is utilized to generate predicted tariff classifications, which correspond to particular tariff rates. In examples, a confidence/accuracy score is applied to the prediction. The predicted tariff codes are then analyzed in view of a set of safeguard rules. If any of the predictions fail the safeguard rules, a manual user review may be triggered. Manually corrected feedback is provided for use in subsequent retraining so that the machine learning model accuracy is improved over time. If any of the predictions pass the evaluation under the safeguard rules, the predictions may be automatically approved.