Research On Intelligent Recommendation System Of Automobile Transaction Based On Knowledge Graph

As the Internet and big data technology are applied more and more widely,people not only obtain massive data,but also have great changes in their lifestyle,among which online shopping has become an indispensable way of shopping.In order to provide users with a good purchase experience,the recommendation system plays an indispensable role.The most representative is the collaborative filtering recommendation algorithm,which is simple and widely used in the industry.However,the algorithm has the problems of cold startup and data sparsity,and lacks the description of data semantic information,so it cannot achieve the accurate description of users or commodities and the accurate recommendation.As a representation of massive amounts of knowledge,knowledge atlas aims to describe various entities and concepts in the real world.Knowledge graph is a kind of semantic network,which can integrate multiple data sources to enrich the semantic information of data and provide services for users with the implied information obtained by inference.As a new analysis tool,knowledge graphing has been widely used in recommendation system.This paper studies the intelligent recommendation system of automobile transaction based on knowledge graph.Taking the automobile trading system as the application background,the domain knowledge graph is constructed by entity and relationship extraction,and the entity,concept and semantic information with high coverage is established.Finally,the traditional collaborative filtering algorithm is integrated with the knowledge graph to realize intelligent recommendation of automobile trading.The research work of this paper has the following parts:(1)study the knowledge extraction method based on web page data,and construct domain knowledge graph.SVM method was used to extract web data entities,and Bootstrap method was used to extract web data from relationships.Taking the transaction data in automobile websites as an example,the effectiveness of the extraction method was verified,and the extracted knowledge graph was stored in the graph database Neo4 j.(2)on the basis of the constructed domain knowledge graph,the article semantic representation of the knowledge graph is carried out.In this paper,the TransE translation model is used to the automobile knowledge graph,and the training model of the objective function is given.The TransE model imparts the constructed knowledge graph into the low-dimensional continuous vector space,while fully retaining the semantic information of the knowledge graph,which provides data analysis guarantee for the following recommendation system.(3)integrate the traditional collaborative filtering algorithm with the knowledge graph analysis tool,and make use of the rich semantic information contained in the knowledge graph to supplement the collaborative filtering algorithm.Experimental results show that the performance of traditional collaborative filtering recommendation algorithm is greatly improved after the fusion.