Research On Personalized Recommendation Algorithm Based On Social Trust And Weighted Sim Rank

The internet information industry of China has achieved a fast development in recent years,and the social networks based internet has greatly enriched people’s daily lives.Users can obtain a lot of information from the internet,however,users can not quickly find their own interest in information from massive data information,facing the problem of information overload.Therefore,the recommendation algorithm based on social network is proposed in order to solve the information overload problem.Recommendation technology can utilize the user’s daily behavior and hobbies to develop personalized recommendations for each user,which can solve the problem of massive data recommendation to a certain extent.However,the recommended quality of traditional collaborative filtering recommendation algorithm is not high due to the sparseness of social network data,which makes cold start users can not get interested information.SimRank is introduced in this paper,and a weighted SimRank algorithm is proposed based on the data of scoring data.Compared with the classical SimRank algorithm,this paper introduces the weighted SimRank algorithm,which is based on the graph structure similarity and it can take full account of the user’s preference for specific projects as well as similarity transmission of the SimRank algorithm,and the F1 value is improved by 32% compared to original SimRank,and the F1 value is improved by 44% compared to traditional collaborative filtring,so this algorithm can alleviate the low recommendation problem of the user’s cold start to a certain extent.In this paper,we achieve MapReduce parallelization of the weighted SimRank algorithm.Compared with the weighted SimRank algorithm,the parallelization algorithm can improve the recommended scalability.Secondly,this paper constructs a reasonable trust network model according to the trust data in social network,and uses the trustworthiness and orientation of trust to calculate the trust between users.The similarity and trust between users are integrated into the traditional collaborative recommendation technique to replace the traditional similarity calculation method in this paper,and the validity of the proposed algorithm is verified by experiments.In summary,the weighted SimRank algorithm and the trust network model are introduced into the traditional collaborative recommendation technology in this paper,and the similarity degree and the trust degree between users are calculated.The experimental results show that the improved algorithm can effectively improve the traditional recommendation algorithm in sparse data set on the recommendation of the low precision.