Machine learning methodology development for autonomous classification of NCM

• Grant number: 17/08629-6
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: May 01, 2018
• End date: January 31, 2019
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Gustavo Peron
• Grantee: Gustavo Peron
• Company: Tradeworks Serviços de Comércio Exterior Ltda
• CNAE: Desenvolvimento de programas de computador sob encomenda
• Pesquisadores principais: Joaquim José Fantin Pereira
• Associated researchers: Tiago Jose de Carvalho
• Associated scholarship(s): 18/08264-0 - Machine learning methodology development for autonomous classification of NCM, BP.TT; 18/08212-0 - Machine learning methodology development for autonomous classification of NCM, BP.TT

Abstract

The objective of the project is to identify and validate the best tools for the development of a methodology based on the machine learning area to achieve future solution of autonomous tax classification for the NCM (Common Nomenclature of Mercosur) classes. This solution is non-existent in the world and will be competitive advantage for the company in its market and segments of activity. NCM is a commodity categorization convention adopted by the Mercosur countries, which is based on the "Harmonized Commodity Description and Coding System" (HS) maintained by the World Customs Organization (www.wcoomd.org). HS is a systematic nomenclature with the following structure: - Ordered List of Positions and Subheadings, with 21 Sections, 96 Chapters and 1,241 Positions, subdivided into Subheadings. The NCM codes are composed of eight digits, the six of which are formed by the HS, while the seventh and eighth codes are specific to the scope of Mercosur. It will indicate the tax rates, the administrative treatments of each product and also will allow the control of imports and exports by the Brazilian government. The classification of goods is one of the most important and controversial issues in an import process, because in many situations there is no specific or predetermined classification. The process typically requires many man hours of expert work from both sides, technical importer and customs specialist. In a market survey carried out by the company, 93% of companies in the automotive segment found products with problems of NCM, with impacts on: increased taxes, fines for improper classification, delays in the release of processes, and risks of process installation Tax and even criminal cases, in case of fraud. In this scenario, we perceive the slowness of the segment's technology providers to respond to systems dedicated to fiscal classification. Thus, we decided to take the lead and invest to develop the best tools for the practice of Foreign Trade. The development of this autonomous process will result in greater efficiency, assertiveness and shorter response time for the overall tax classification process. In order to achieve this goal, it will be necessary to go through an intermediary stage to apply the "machine learning" that will enable and compose the creation of an 'artificial intelligence' to classify the items that may result in greater efficiency of the process as a whole . Thus, this project will be characterized by the improvement of a fiscal classification process associated with the research and development of an algorithm of machine learning and artificial intelligence, presented in the form of computational software. As a result of this work, it is planned to develop this unique and innovative application that will provide support for decision making, notably aimed at the fiscal classification of items that allows a fast framing capability and with great assertiveness. (AU)