Research On Material Coding Text Recognition Based On Machine Learning

At present,enterprises are gradually pursuing high-quality and efficient development.They implement the concept of refined management to further improve the management level.For some large manufacturing enterprises and enterprises with a large number of materials storage and docking with many suppliers,it is more necessary to transfer their logistics and warehousing business to the third-party logistics companies,so as to transfer their energy to the core business.However,for the third-party logistics enterprises,it is difficult to fully grasp the material data information of the demander and multiple suppliers,and the materials provided by multiple manufacturers have the problems of inconsistent name and incomplete information,and the list of suppliers and the materials they supply are constantly changing,which makes it difficult for the third-party logistics enterprises to accurately classify the materials of different suppliers into a unique category Coding reduces the efficiency of material flow,increases the cost of material management and personnel training,and hinders the cooperation of enterprises in the supply chain.Therefore,it is of practical significance and necessary to establish a standardized and high accuracy coding matching recognition model.Based on this problem,considering that the material name under the same category code has strong relationship between keywords and semantics,this paper takes the material supply data of Z Company as the original data,establishes the code recognition model based on text analysis technology in the field of artificial intelligence,and through data cleaning,manually establishes the code material corresponding table with correct corresponding relationship,taking the category code as the reference Tag,semantic analysis and large sample training are carried out for the text keywords of many material names and specifications under each tag,so that the model can form semantic and keyword recognition rules for each category coding,thus giving the recognition reference results for the unique category coding of new materials.The main work of this paper includes: designing the material data preprocessing module,realizing the material code splitting,the most detailed category code recognition,word segmentation,sample and classification In this paper,the functions of tag digitization,word2 vec word vector space model and material coding recognition machine learning model are introduced,including traditional machine learning model and deep neural network model,and the model is trained.Through repeated experiments and adjustments of model structure and parameters,the optimal structure and parameters of each model suitable for material coding recognition are obtained The results show that the application of machine learning model in this scenario is feasible,and under this data background,the best algorithm used by traditional machine learning is logical regression,and the accuracy rate of material code recognition is 90.62%;the best deep neural network learning model is bidirectional LSTM model,and the accuracy rate is 90.73%?...