Classification methods and models for automatic determination of goods code by foreign economic activity goods nomenclature

Abstract

Classification methods and models for automatic determination of goods code according to the commodity nomenclature of foreign economic activity (CN FEA) are considered. In the classification procedure, the object is reduced to one or more classes based on the results of the comparison to evaluate the proximity and make a conclusion. The relationship between vectors calculated by scalar multiplication was used as a measure of proximity. In order to determine the most informative symbols for classes, linguostatistical methods based on the information about the a priori probability of occurrence of terms and symbols were used. In doing so, macro-average and micro-average methods, which are considered effective in assessing the quality of classification for several classes, were used. Also, on the basis of syntactic and linguostatistical analysis, a generalized scheme of the process of automatic classification of goods in column 31_1 of the goods declaration graph is proposed. This recommended method of automatic identification of the CN FEA code allows participants of foreign economic activity to ease the processes of filling out declarations for goods.