Research On Movie Recommendation Algorithm Integrating Knowledge Graph And Attention Mechanism

With the development of Internet technology,The data volume of network users and items has increased exponentially,so the problem of information overload has become increasingly serious,and the recommendation system has emerged as the times require.However,traditional recommendation algorithms face serious data sparseness,difficulty in capturing high-order relationships between users and items,dynamic changes in user interests and high complexity of recommendation models.In order to solve the above problems,this thesis combines the knowledge graph and deep learning methods to carry out the following research in the field of movie recommendation,and the main research contents are as follows:1.Aiming at the sparse data in the field of movie recommendation,the inability to fully capture the high-order relationship between users and items,and the dynamic changes of user interests,a movie recommendation algorithm that integrates knowledge graphs and convolutional attention networks is proposed.Firstly,the knowledge graph is introduced as auxiliary information to better capture the high-order relationship between users and movies through the connection of nodes,thereby alleviating the problem of data sparsity.Secondly,a convolutional neural network is introduced to process the embedded high-dimensional knowledge graph triples into low-dimensional continuous vectors to obtain the characteristics of users and movies.Then,an attention mechanism module with dynamic factors is proposed to find out the short-term dynamic interest changes of users by only referring to the movies recently interacted by users.Finally,the similarity scores obtained by various candidate movies are sorted in descending order to achieve the final TOP-K movies recommendation.Through simulation verification,the algorithm has at least 4.2% and 4.3% improvement in recommendation accuracy and recall rate,and the movie recommendation performance has been significantly improved.2.Aiming at the sparse data in movie recommendation algorithms,the inability to fully capture the high-order relationship between users and items,and high complexity of recommendation models,a movie recommendation algorithm that integrates collaborative knowledge graphs and optimized graph attention networks is proposed.Firstly,the single knowledge graph is combined with the interaction information between users and movies to become a collaborative knowledge graph.It not only strengthens the relevance and interaction between users and movies,but also alleviates the problem of data sparsity to a greater extent.Secondly,an optimized graph convolutional network is introduced,which greatly reduces the model complexity without affecting the overall recommendation performance by removing feature transformation and nonlinear activation modules based on the traditional graph convolutional network.Then,the optimized attention mechanism module is used to capture the deviation between the recommended movies and the movies of interest to the user in time,and automatically assign distinct weights.Finally,the similarity scores obtained by distinct candidate movies are sorted in descending order to achieve TOP-K movies recommendation.Through simulation verification,the algorithm has at least2.8% improvement in movie recommendation accuracy,and 56% optimization in time complexity.