Research On Recommendation Algorithm Fusion Of Recurrent Knowledge Graph And Collaborative Filtering

With the continuous development of the film industry,film culture has gradually become an important content of people’s daily cultural needs.However,in the face of massive amounts of data,how to quickly and accurately obtain the desired information has become a huge challenge.In this context,the movie recommendation system was born,the core of which is the recommendation algorithm.Most of the existing algorithms are based on the traditional collaborative filtering recommendation algorithm.This algorithm is currently the most successful and widely used,but there are still problems such as data sparsity and poor accuracy.Aiming at the problems of existing recommendation technology,this paper proposes a recommendation algorithm(RKG-BCF)that combines Recurrent Knowledge Graph Recommendation Algorithm and Collaborative Filtering(Collaborative Filtering,CF)by studying movie recommendation algorithms.The algorithm is a fusion of knowledge map and deep learning model(RKG)and improved collaborative filtering recommendation algorithm(BCF).The main work is as follows:(1)In view of the data sparseness and cold start problems faced by the current movie recommendation algorithm,this paper studies the knowledge graph in the movie domain,and constructs a knowledge graph that contains rich user-movie entities and relationship information between entities.Using the movie knowledge map construction technology,the user’s historical behavior and extracted movie knowledge are integrated into the knowledge map,which provides richer and more accurate information for the recommendation system,and is stored in the graph database Neo4 j for easy query,more efficient management and The use of knowledge map information alleviates data sparseness and cold start problems to a certain extent.(2)In order to further improve the accuracy and interpretability of the collaborative filtering algorithm,this paper studies the traditional user-based collaborative filtering recommendation algorithm,taking into account the strict factors of movie quality and users themselves,and using two independent biases to represent,to improve the accuracy and interpretability of recommendations.In the end,there was a significant improvement in the comparative experiment.(3)In view of the fact that the traditional movie recommendation fails to deeply tap the potential interests of users,this paper studies the fusion model of knowledge graph and recurrent neural network,and extracts the semantics between users and movies through the constructed knowledge graph in the movie domain through the Trans H knowledge representation method Path,and the semantic path is input into the recurrent neural network for training,and finally the prediction score is generated through the activation function,and the result of movie recommendation is obtained.Experiments were carried out on the Movicel Lens-1M movie dataset.In the comparison results of the comparison algorithms,it was found that the actual effect of the algorithm in this paper has been significantly improved,and the accuracy of the recommendation has been effectively improved.To sum up,through the research on the current movie recommendation algorithm,this paper combines the traditional collaborative filtering algorithm and the algorithm combined with knowledge map and deep learning for improvement,and conducts experimental comparisons.The results show that the performance of the RKG-BCF algorithm has obvious advantages.Improvement,effectively improving the accuracy and interpretability of movie recommendations,and alleviating data sparsity and cold start problems.