Font Size: a A A

Research On Personalized Recommendation Algorithm Based On Trust Mechanism

Posted on:2017-08-10Degree:MasterType:Thesis
Country:ChinaCandidate:Y TianFull Text:PDF
GTID:2358330503488907Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the high-speed development of the Internet, the amount of network information has increased dramatically, users can not quickly and effectively get their own useful information in the face of massive data information, which is the problem of “information overload”. Search engine and recommender system is an important technical means to solve the problem of “information overload”, but with the rapid growth of the amount of information, search engine of passive search has not fully satisfy the people's needs, personalized recommender system has been widely studied and developed because it can take the initiative to recommend information which can satisfy users' interests. Personalized recommendation collects and analyzes the historical behavior of the user, and take the initiative to provide users with personalized recommendation using a certain algorithm mechanism. Collaborative filtering recommendation algorithm is the most widely used algorithms in recommender systems, but there are some problems in the traditional collaborative filtering recommendation algorithm, such as data sparsity and cold start problems and so on, these problems led the recommended quality is not high for cold start users and items, the algorithm needs to be improved. With the rise and development of social networks, people are not only the consumers of information, but also the producer of information, the information have a rapid growth, therefore, some researchers have proposed to join people's social information to recommend for users. Trust information was introduced to recommender system as the most important social information, and a trust based recommendation algorithm was proposed, it is based on the traditional algorithm, and improve recommendation quality by adding the use's trust information and effectively alleviate the problems existing in traditional collaborative filtering recommendation algorithm, but there are some problems such as data sparsity and single value form in the trust information.In view of the problems of trust recommendation, this paper proposes the following three algorithms based on the analysis of the traditional collaborative filtering recommendation algorithm and trust recommendation algorithm:(1) For the problem that explicit trust information is sparse and monotonous, we introduced implicit trust inference algorithm, and combined explicit trust and implicit trust, and put forward the EITrustSVD algorithm based on double trust mechanism that based SVD++ model, we can get a reliable recommendation depending on explicit trust, and get a recommendation related to user preferences relying on implicit trust at the same time. It show that this algorithm effectively improves the recommendation accuracy in the experiments on publicly available Film Trust data sets.(2) For the case that trust information is trust list or binary trust, we combined the rating information and trust relationship to study the trust value between users, then we took this trust value as a recommended weight, and put forward a collaborative filtering recommendation algorithm with trust learning, and experiments on publicly available Ciao data sets show that the proposed algorithm can effectively provide the recommendation accuracy.(3) By means of analyzing the relation between the trust list of target users and the similarity of these users, we selected effective recommended users in target user's trust list and combine with the dynamic changes of the similarity, then a trust recommendation algorithm with dynamic similarity was put forward.
Keywords/Search Tags:Collaborative Filtering, Trust Recommendation, Explicit Trust, Implicit Trust, Trust Learning, Dynamic Change
PDF Full Text Request
Related items