Font Size: a A A

The Research On Collaborative Filtering Recommendation Algorithm Based On Trust Relationship

Posted on:2019-05-01Degree:MasterType:Thesis
Country:ChinaCandidate:Y S ZhangFull Text:PDF
GTID:2428330545969674Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
As an effective information filtering method,personalized recommendation has been widely used in e-commerce,video sharing,online social networking and other fields.Collaborative filtering is the most commonly used type of algorithm in the current recommendation system.The basic idea of it is to analyze a large number of users' “scores” of items to find the similarities of user preferences,and then recommend the items that meet their preferences to users.Although the collaborative filtering algorithm solves the problem of information redundancy to some extent,there are still some problems such as cold start,data sparsity,and so on.The development of social networks brings new opportunities for the research of recommendation algorithms.Scholars introduce trust information into the recommendation algorithms.By adding trust information,the problems of data sparseness and cold start can be better mitigated,and better recommendations are obtained.This article studies collaborative filtering algorithms based on social trust relationships.The main tasks are as follows:First,a collaborative filtering algorithm based on dual trust mechanism is proposed.In order to solve the problem of data sparse and cold start in the traditional collaborative filtering algorithms,the social trust recommendation mechanism is introduced into the recommendation system.By adding the user's explicit trust information,the above problems can be effectively mitigated.However,explicit trust is more difficult to obtain,and data is sparse.In order to better improve the efficiency of recommendation,this paper adds implicit trust information based on the recommendation algorithm based on display trust and obtains reliable recommendation through explicit trust.Receive recommendations related to user preferences through the influence of implicit trust.Experimental results show that the new algorithm can effectively improve the accuracy,coverage and overall performance of the recommendation.Second,a trust recommendation algorithm based on improved similarity is proposed.The recommendation algorithm based on the social trust relationship needs to integrate the scoring data of the trusted neighbor to obtain a new score,and the merged score is likely to have misaligned noise data.In this paper,fusion score coefficients are proposed to ensure the correctness and reliability of fusion scores from trusted neighbors.On the basis of classical similarity measurement methods,the modified Jaccard coefficients and fusion score coefficients are used as weight coefficients to compare the traditional similarity.The calculation results are improved.Experimental results show that the improved algorithm can effectively improve the prediction accuracy.
Keywords/Search Tags:Recommender Systems, Collaborative Filtering, Social Trust, Data Sparsity and Cold-Start, Trust Propagation
PDF Full Text Request
Related items