Intelligent Classification Algorithm Research And System Implementation Of Customs Commodity

With the continuous expansion of China's import and export trade,the classification of customs commodity has become an important issue.The traditional commodity classification process is completed by professional categorizer.When a large number of commodities need to be classified,there are problems of time-consuming,high cost and low efficiency.Intelligent classification can improve customs clearance efficiency,reduce transaction risk and save a lot of manpower and material costs.Focusing on the actual project requirements,this thesis conducts research on the customs classification algorithm of customs commodity based on the special scenarios of customs commodity classification and on big data,natural language processing,machine learning and deep learning.This thesis proposes a customs commodity hierarchy intelligent classification algorithm,which is based on index and machine learning method.The judgment of different digit is carried out according to the level of chapter,section,head and sub-head.The higher accuracy of algorithm is proved by theoretical analysis and experiment;Finally,the customs intelligent classification system is designed and implemented,and the web interface and API are provided to meet the user's use requirements.The main work of this thesis includes the following aspects:(1)Propose a dictionary and feature construction method in the customs commodity field.To solve the word segmentation problem in the field of customs commodity through constructing a domain dictionary.By selecting the appropriate feature type and feature dimension,a better feature representation is obtained.(2)Design the intelligent hierarchy classification algorithm based on index and machine learning.The algorithm derives the customs commodity code by step-by-step analysis,obtains the four-digit code by establishing and searching the inverted index and the four-digit code binary classifier,obtains six-digit code through the four-digit to sixdigit code classifier and the relevant strong rule of the classification constraints then obtains the ten-digit code through six-digit to ten-digit code classifier.Comparing experimental results with traditional machine learning algorithms,fast Text text classification model,GCN graph convolutional neural network algorithm and the advanced algorithm in the customs commodity classification field.The results show that the algorithm has the highest classification accuracy.(3)Design and implement the customs commodity classification system.Design the classification system architecture and user interaction logic,optimize the storage and computing logic,improve the system response speed,provide the web interface and API for users,deploy the cloud service online finally.The customs commodity hierarchy intelligent classification algorithm designed in this thesis can be used in many classification scenarios.The experiment proves that the algorithm has high classification accuracy and can meet the needs of users and help relevant users to complete the classification process quickly.