Research On Recommendation Algorithm Based On User Indirect Trust And Behavior Sorting

Under the impact of the modem Internet and big data,the data scale grows rapidly,and the user demand tends to be diverse.The traditional technical means can't handle the overload problem caused by massive data.As an important branch of data mining,the recommendation algorithm has the advantage of alleviating data redundancy in the face of large-scale data processing.It can not only extract the data that users are interested in,but also push the valuable long tail information to the user.Compared with traditional recommendation,the recommendation algorithm based on user trust introduces trust assistance data to alleviate data sparseness and improve the performance of the algorithm to some extent.However,most studies do not deeply analyze the trust correlation between users,and do not consider the dynamic changes of trust will affect the accuracy of similar users.To this end,this paper proposes a dynamic computing model of indirect trust,and designs two recommendation algorithms based on user indirect trust.The main work of this paper is as follows:(1)Propose a dynamic calculation model of indirect trust.The model mainly uses Newton's law of cooling,incorporates the time decay factor for indirect trust,and changes the difference value of the trust before and after,to adjust the deviation caused by data sparseness by similar users.(2)Design a recommendation algorithm based on user dynamic trust and Wilson sort.The algorithm uses the indirect trust dynamic calculation model to calculate the trust change,and introduces the similarity calculation from the trust change value.Secondly,the Wilson sort algorithm is used to intervene in the Top-N list for recommendation.Experiment results verify effectiveness of the algorithm.(3)Design a recommendation algorithm based on user indirect trust and Gaussian padding.The algorithm first analyzes the user trust transfer network graph.Since the indirect trust data cannot be directly collected,the ratio of each branch transfer node to the total path node in the network graph is used,and then the entire trust indirect value is obtained by multiplying one by one;secondly,the Gaussian function is obtained.Pre-populate the user scoring matrix to reduce the error of similar calculations;finally,the calculated similarity is combined with the user's trust to make recommendations on the neighbor prediction.Experimental results show that the algorithm improves the accuracy of the recommendation.All in all,this paper starts from the work of trust recommendation algorithm,and further explores the user trust relationship and improves the recommendation accuracy.The indirect and dynamic trust are combined with Gaussian function and Wilson order to design a new recommendation algorithm.The RMSE and MAE evaluation indicators verify the effectiveness of recommendation algorithm on the recognized FilmTrust dataset.