Research On Key Technologies Of Personalized Movie Recommendation Based On Knowledge Grap

With the continuous progress of Internet technology,we have now entered the era of big data.How to quickly obtain the desired information from the vast amount of data has always been an important issue.The emergence of recommendation system solves the information acquisition dilemma greatly and is applied in every aspect of life.However,the traditional recommendation system inevitably has the problems of sparse data and cold startup.The optimization of recommendation system is a hot topic for many scholars,and the recommendation of fusion knowledge graph is one of the important research directions.As a structural network with rich semantic information,knowledge graph significantly enhances the connection between users or items,which has a good effect on solving a series of problems in the recommendation system.This paper focuses on the field of film recommendation and studies the personalized film recommendation based on the fusion of knowledge graph.The main work results are as follows:(1)The knowledge map of the film field is constructed.Compared with the previous knowledge map of the film,this paper innovatively introduces the user’s emotion score of the film into the knowledge map,enriching the semantic connection between the films and enriching the attribute information of the films.On the basis of the overall analysis of the construction of film knowledge map,the construction of pattern layer and data layer is completed.(2)TranSparseT,a knowledge representation model based on translation,is established.For the knowledge map of the film domain,the number of head and tail entities of genres and films is inconsistent and greatly different,which leads to the unbalanced problem of other knowledge representation models in this special application field.This paper proposes that using TranSparseT model can effectively solve the above problems.The knowledge representation model is trained based on the knowledge map of the film domain,and the link prediction is taken as the validation index.The optimal parameter group of the model is obtained through continuous training and optimization.(3)The TranSparseT-CF personalized recommendation algorithm is proposed,which combines the semantic similarity of the film based on the knowledge graph with the item similarity based on the user’s comprehensive score to select the appropriate ratio and get the final similarity of the film.Through experiments,the optimal fusion factor is determined,and then the score prediction and Top-N recommendation are carried out.The accuracy and recall rate are taken as evaluation criteria.Through experiments,the recommendation results of the traditional collaborative filtering recommendation algorithm and other translation model based recommendation algorithms are compared,and the results prove that the personalized recommendation algorithm proposed in this paper has better recommendation results.To sum up,the TranSparseT-CF personalized recommendation algorithm proposed in this paper can effectively improve the recommendation effect of movies by integrating the knowledge graph into the recommendation process and taking advantage of the influencing factors of emotional score.