Research On The Recommendation Algorithm Based On Word Vector

With the rapid development of Internet information technology,more and more customers rely on online platform to purchase goods and get information.However,with the advent of the digital age and the further increase in the amount of information,people are more and more difficult to identify and obtain effective information in the face of a wide variety of goods in the network mall.The recommendation system came into being.As a means of product filtering and user requirement positioning,the recommendation system can arrange and recommend the product according to the consumer's needs and historical behavior data,and quickly locate the product that the user needs,thus helping the user to save the time and energy wasted by browsing a large number of useless information and effectively solve the problem of information overload.This paper constructs recommendation algorithm framework and user-commodity similarity model based on word vector model,so as to generate recommendations.The specific work is as follows.First,this paper first analyzed and studied the two commonly used recommendation algorithms and their recommendation principles in the recommendation system,and reviewed the related research progress on the recommendation algorithms at home and abroad,and summarized the current situation at home and abroad,to provide the theoretical reference for the further research of the recommendation algorithm.Second,built the recommendation algorithm framework and the model of usercommodity similarity based on word vector model.This paper described the theory of word similarity based on word vector model,and the advantage and value of word vector model in similarity recommendation are explained by comparing vector space model.Finally,the word vector model was applied to the content-based recommen-dation to build the recommendation algorithm framework and the user-commodity similarity model.Third,this paper compared the recommendation algorithms based on user preferences and product characteristics by contrast experiment.Based on customers' order data,this paper firstly to extract the features of the goods,then to study the preference characteristics based on the customers' history purchase data,and then took the extension of the word vector-doc2 vec method to calculate the similarity between the user preferences and the characteristics of the goods in the experiment.The experimental results prove the advantages of word vector algorithm in text similarity operation for context word order and semantic computing,and the feasibility of applying it to the recommendation algorithm.