<abstract xmlns="http://www.w3.org/1999/xhtml"><p>Every five years, the U.S. Census Bureau conducts the Economic Census, the official count of US businesses and the most extensive collection of data related to business activity. Businesses, policymakers, governments and communities use Economic Census data for economic development, business decisions, and strategic planning. The Economic Census provides key inputs for economic measures such as the Gross Domestic Product and the Producer Price Index. The Economic Census requires businesses to fill out a lengthy questionnaire, including an extended section about the goods and services provided by the business.</p><p>To address the challenges of high respondent burden and low survey response rates, we devised a strategy to automatically classify goods and services based on product information provided by the business. We asked several businesses to provide a spreadsheet containing Universal Product Codes and associated text descriptions for the products they sell. We then used natural language processing to classify the products according to the North American Product Classification System. This novel strategy classified text with very high accuracy rates - our best algorithms surpassed over 90%.</p></abstract>

Every five years, the U.S. Census Bureau conducts the Economic Census, the official count of US businesses and the most extensive collection of data related to business activity. Businesses, policymakers, governments and communities use Economic Census data for economic development, business decisions, and strategic planning. The Economic Census provides key inputs for economic measures such as the Gross Domestic Product and the Producer Price Index. The Economic Census requires businesses to fill out a lengthy questionnaire, including an extended section about the goods and services provided by the business.To address the challenges of high respondent burden and low survey response rates, we devised a strategy to automatically classify goods and services based on product information provided by the business. We asked several businesses to provide a spreadsheet containing Universal Product Codes and associated text descriptions for the products they sell. We then used natural language processing to classify the products according to the North American Product Classification System. This novel strategy classified text with very high accuracy rates - our best algorithms surpassed over 90%.

Applying Machine Learning for Automatic Product Categorization

Journal of Official Statistics

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

{"article-title":"Applying Machine Learning for Automatic Product Categorization"}

Every five years, the U.S. Census Bureau conducts the Economic Census, the official count of US businesses and the most extensive collection of data related to...

Applying Machine Learning for Automatic Product Categorization

Pubblicato online: 22 giu 2021

Pagine: 395 - 410

Ricevuto: 01 mag 2019

Accettato: 01 mar 2020

DOI: https://doi.org/10.2478/jos-2021-0017

Parole chiave
Text analytics, artificial intelligence, data collection

© 2021 Andrea Roberson, published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Applying Machine Learning for Automatic Product Categorization

Pubblicato online: 22 giu 2021

Pagine: 395 - 410

Ricevuto: 01 mag 2019

Accettato: 01 mar 2020

DOI: https://doi.org/10.2478/jos-2021-0017

Parole chiaveText analytics, artificial intelligence, data collection

© 2021 Andrea Roberson, published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Parole chiave
Text analytics, artificial intelligence, data collection