A Movie Recommendation Algorithm Based On Trust Relationship

Since entering the era of big data,the personalized recommendation algorithm has received extensive attention,not only solves the problem of information overload,but also mines a lot of useful information from massive data.In addition,the development of social platforms has brought people together,and more and more social platforms have integrated social attributes.For a user,the recommendation of his friend may be more interested and trustworthy.Therefore,the social relationship between users has a significant influence on the accuracy of the recommendation algorithm,and the user's social trust and personalized recommendation are integrated into one.It is necessary and meaningful to study.In order to alleviate the impact of data sparseness on recommendation accuracy in the traditional collaborative filtering recommendation algorithm,the trust relationship is incorporated into the movie recommendation,based on the user's trust in the movie rating data and trust in the social information.Through the dynamic adaptive integrated trust weights for linear fusion,combined with the probability matrix decomposition model,a Multiple Trust-Probability Matrix Factorize(MT-PMF)algorithm is proposed.The algorithm is validated on the film dataset FilmTrust and compared with related algorithms.The results show that the accuracy of MT-PMF scoring prediction is effectively improved on both MAE and RMSE indicators.The main research contents of this paper are as follows:(1)According to the similarity between the user's historical score data,the target user similarity mean is used as the threshold for selecting similar users,and the scores of the similar users on the historical common score items are used to predict the target user's scores on the items.The score trust degree between the target user and the similar interest user is determined by judging the deviation between the predicted score and the actual score.The advantage of this scoring trust calculation method is that it can deeply analyze the reliability of users with similar scores and improve the accuracy of predictive scores.(2)Since the trust relationship existing in the user's social interaction tends to affect the preference of the movie,considering that the trust between users is weaker in the process of transmission,the user's trust to the direct friend is higher,and the trust of the friend of the friend is second.Therefore,in order to reduce the complexity of the algorithm,improve the time efficiency,and effectively retain the most valuable trust users,this paper only calculates the social trust between users by considering the two-level trust user relationship.This method is simple and easy to understand,and can alleviate the problem of low recommendation accuracy caused by sparse scoring data.And in practical applications,the algorithm is more explanatory.(3)The score trust and social trust obtained above are dynamically and linearly integrated to obtain the comprehensive trust between users.The user's behavior is directly affected by users who believe that the overall trust is high.The MT-PMF algorithm is designed by using the integrated trust weight as a regularization term for the probabilistic matrix factorization.The algorithm performs validation on the FilmTrust dataset.The algorithm alleviates the impact of data sparseness,and improves the accuracy of scoring prediction.The algorithm is more scalable.