Parallel LSTM structure customs commodity classification method based on corresponding degree measurement

Abstract

The invention discloses a parallel LSTM structure customs commodity classification method based on corresponding degree measurement, and relates to the field of customs tax administration. The methodis composed of a parallel network structure of different feature data and two network output part feature corresponding mechanisms, including digital vector generation of text data, responsivity measurement of deep parallel LSTM, and an algorithm structure which is suitable for softmax probabilistic definition confidence coefficient to specially process diversification of classification elements.According to the method, through a parallel network structure of different data and a plurality of network output part feature corresponding mechanisms, priori knowledge and data processing are carried out on customs commodities, and a big data technology and deep learning calculation are utilized to use customs mass data for model optimization, so that the accuracy and the accuracy of the customstax rule number are improved.

Description

technical field [0001] The invention belongs to the field of customs taxation, and in particular relates to a parallel LSTM structure customs commodity classification method based on corresponding degree measurement. Background technique [0002] Customs declaration goods have to pay different proportions of taxes and fees. These taxes and fees are uniquely determined by the tariff number. However, in actual production and life, because the company itself does not have a thorough grasp of the classification knowledge, the daily customs declaration data is very large, and the customs declaration tariff Insufficient verification personnel and other circumstances make it difficult to collect national customs duties and fees accurately. [0003] In order to solve this problem and liberate a lot of manpower and material resources, we have designed a customs classification algorithm based on LSTM. [0004] In the actual algorithm design, we encountered many technical difficulties...

AI Technical Summary

Problems solved by technology

[0013] The purpose of the present invention is to provide a parallel LSTM structure customs commodity classification method based on corresponding degree measurement, by performing prior knowledge and data processing on customs commodities, using big data technology and deep learning calculations to use massive customs data for model optimization, It solves the problems of difficult management and inaccurate statistics of existing customs tariff codes

Images