RobotRobot
Home/Success Stories /

Commercial goods valuation and classification automation

Commercial goods valuation and classification automation

Natural Language Processing (NLP)

Statistics

Highlights

Challenge

Automate the process of classification and valuation of commercial goods

Solution

In-house-built ML models for the classification and valuation of commercial goods based on commercial description and other descriptive fields.

Results

Automation of over 80% of the work with more than 95% accuracy.
bg

About the Project

The Client was a commercial goods classification and valuation solution and service provider to various Customs offices in the world. The Client was looking for an ML solution that would help their employees identify the applicable commodity codes for the goods based on the provided commercial descriptions then use the commodity code, descriptive information on the item, and external data sources to predict a reasonable price for the goods. The requirement was to build a solution that would be easy to customize and integrate into various systems in different countries. To solve this problem, the Portmind team used Natural Language Processing and Statistics techniques and automated the repetitive, time-consuming, and error-prone tasks of manually classifying and valuating commercial goods.

CHALLENGE

Predict HS codes and prices for commercial goods

International trade is a fundamental aspect of the global economy. With goods moving across the border, governments around the world are able to collect taxes. The goods are typically taxed based on the type of the product that is identified according to the HS code classification system that defines product categories. Considering that there are thousands of categories, it is challenging to identify and validate the correct classification for a product. Moreover, importers tend to misclassify the goods by assigning an incorrect category to pay less taxes.

To prevent this from happening, valuation and classification analysts working at the Client company check each item’s information within a trade transaction to make sure that the declared classification code and price are correct. This process involves many manual steps, like using classification books to find the appropriate classification code for the goods, checking the price against the past transaction data or finding the price on the market, etc. These processes are highly repetitive, time-consuming, error-prone, and subject to human bias.

    SOLUTION

    Based on the commercial descriptions of the goods in the format of free text, the ML solution predicts applicable HS codes per product. This information, combined with the country of origin, unit of measurement, and other descriptive fields of the item, is used to predict reasonable price ranges for products. In addition to price range prediction, the ML solution provides a list of transactions with similar goods from the historical data and international data of prices, further assisting officers to make data-driven decisions.

    Based on the Clients’ business needs, we delivered a product that does the following actions:

    Predicts the applicable HS code based on the commercial goods description together with the confidence of prediction
    Provides a reasonable price-range for the item
    Retrieves similar transactions from the Client’s historical data, helping users in further decision making
    Retrieves price evidence from various external data sources, assisting users in further decision making

    RESULTS

    The Portmind team developed a solution that automates on average over 80% of the work of classification and valuation analysts with over 95% accuracy. With these solutions in place, the Client was able to save resources, increase staff efficiency and customer satisfaction, minimizing the risk of misclassification and undervaluation for governments around the world.

    Want to see how the solution works in practice?