Application Research Of Recommendation Algorithm Based On Knowledge Graph And Deep Learning

The development of Internet technology makes the amount of data increase rapidly,which makes it very troublesome for people to select what they are interested in from the huge amount of information.In order to improve user experience and meet the personalized needs of users,recommendation system emerged and gradually became an indispensable existence in people's life.The commonly used traditional recommendation algorithm is relatively simple,and the recommendation accuracy needs to be improved.Although the deep learning recommendation algorithm has optimized the final recommendation effect,it has a little insufficient interpretation of the results.Auxiliary information can enrich the semantic features of users or items in the recommendation system,and the combination of auxiliary information and recommendation system has become the optimization direction of many scholars' in-depth research.Knowledge Graph as a kind of semantic network,can learn different potential relationship between for the project.Combining with the Knowledge Graph,this thesis proposes a recommendation method based on cross attention unit.Through the newly designed feature cross module,the model can learn more association information between the items to be recommended and the corresponding entities in the Knowledge Graph.In this thesis,comparative experiments are carried out on the datasets of the recommendation field,and the results verify that the proposed model and feature cross module can effectively carry out feature interaction between items and entities,so as to enhance the recommendation effect.After analyzing relevant requirements,this thesis constructs a movie recommendation system by using Python and related frameworks.The research work of this thesis is mainly divided into the following:(1)This thesis describes the research status of the mix of Knowledge Graph and recommendation algorithm,summarizes relevant ideas and techniques,including traditional recommendation algorithm,deep learning recommendation algorithm and knowledge graph recommendation algorithm,and explains the related construction process of Knowledge Graph.(2)In view of the advantages of Knowledge Graph and the fact that most recommendation models fail to make full use of the association between Knowledge Graph and recommendation items,this thesis proposes a Knowledge Graph recommendation algorithm(CAKR)based on cross-attention fusion.CAKR model contains cross-attention fusion module,recommendation module and Knowledge Graph embedding module,the input of two task modules is learned alternately in low-dimensional space by cross-attention fusion,to interact with the two embedding vectors of the item and entity.Then the feature vectors are sent into the relevant task modules respectively,and the final results are calculated by their respective prediction functions.In this thesis,comparative experiments are conducted in three common datasets,and the results verify that the CAKR model can effectively improve the recommendation effect.(3)Based on the proposed recommendation model,this thesis constructs a movie recommendation system,which is divided into three core layers,including data source processing,user interaction layer and recommendation engine.The front and back end involves Dangjo,MySQL,HTML,CSS3,jQuery and other related technologies.