Methods and systems for prediction of tariff classifications

A method, comprising:

• receiving, from one or more databases, historical data associated with each of a plurality of historical items, the historical data comprising:
• one or more characteristics;
• and
• an assigned tariff classification;
• providing the historical data to a machine learning platform;
• training the machine learning platform based on the historical data to generate a trained machine learning model for predicting item tariff classifications, the trained machine learning model comprising:
• a text processing layer configured to regularize received item descriptions;
• a pretrained language processing model configured to generate a plurality of embeddings;
• a first dense layer configured to dimensionally change the plurality of
• embeddings to create dimensionally changed embeddings and provide the dimensionally changed embeddings as repeated embeddings to a bidirectional long short-term memory layer;
• the bidirectional long short-term memory layer, configured to process the repeated embeddings in a forward and a backward direction;
• a second dense layer configured to dimensionally change the processed repeated embeddings of the bidirectional LSTM layer and output the dimensionally changed processed repeated embeddings to a tariff classification prediction layer;
• and
• the tariff classification prediction layer;
• receiving item data associated with a new item, the item data comprising one or more characteristics;
• providing the item data to the trained machine learning model;
• predicting, at the trained machine learning model, a predicted tariff classification of the new item;
• applying one or more rules to the predicted tariff classification;
• based on the application of the one or more rules, automatically determining whether to approve the predicted tariff classification;
• receiving, from the one or more database, updated historical data associated with each of an updated plurality of items, the updated plurality of items comprising the new item, the updated historical data comprising the one or more characteristics of the new item and the predicted tariff classification of the new item;
• and
• retraining the machine learning platform, using the updated historical data, to generate a retrained updated machine learning model for predicting item tariff classifications.

A computing system, comprising:

• at least one processor;
• and
• at least one memory storing computer-executable instructions, the computer-executable instructions when executed by the at least one processor causing the computer to:
• receive, from one or more databases, historical data associated with each of a plurality of historical items, the historical data comprising:
• one or more characteristics;
• and
• an assigned tariff classification;
• provide the historical data to a machine learning platform;
• train the machine learning platform based on the historical data to generate a trained machine learning model for predicting item tariff classifications, the trained machine learning model comprising:
• a text processing layer configured to regularize received item descriptions;
• a pretrained language processing model configured to generate a plurality of embeddings;
• a first dense layer configured to dimensionally change the plurality of embeddings to create dimensionally changed embeddings and provide the dimensionally changed embeddings as repeated embeddings to a bidirectional long short-term memory layer;
• the bidirectional long short-term memory layer, configured to process the repeated embeddings in a forward and a backward direction;
• a second dense layer configured to dimensionally change the processed repeated embeddings of the bidirectional LSTM layer and output the dimensionally changed processed repeated embeddings to a tariff classification prediction layer;
• and
• the tariff classification prediction layer;
• receive item data associated with a new item, the item data comprising one or more characteristics;
• provide the item data to the trained machine learning model;
• predict, at the trained machine learning model, a predicted tariff classification of the new item;
• store the predicted tariff classification in the one or more databases;
• receive, from the one or more database, updated historical data associated with each of an updated plurality of items, the updated plurality of items comprising the new item, the updated historical data comprising the one or more characteristics of the new item and a retraining tariff classification of the new item;
• provide the updated historical data to the machine learning platform;
• and
• automatically retrain the machine learning platform based on the updated historical data to generate a retrained updated machine learning model for predicting item tariff classifications.