The Research Of Personalized Recommendation Algorithm Combining Trust

Current research in personalized recommendation also has many challenges, such as problems about data sparseness and cold start,especially personalized recommendation in the social network environment has many to explore, so the research of personalized recommendation combining trust in the social network environment is meaningful.To solve the problems, the traditional recommendation methods are analyzed, including the classification of the traditional recommendation methods and the problems that faced;then the research of personalized recommendation combining trust are analyzed.Through the above analysis, the thesis puts forward the following three methods of personalized recommendation combining trust:the thesis proposes a collaborative filtering recommendation algorithm based on trust propagation, algorithm proposes TSR weight combing trust, similarity and relationship to replace the similarity in traditional collaborative filtering algorithm in order to find neighbors TSRCF algorithm solves the data sparse problem and helps users getting high quality recommendations in the information overload environment. Experimental results based on Epinions data sets and FilmTrust data sets demonstrate that the algorithm performs better than the traditional filtering algorithm in terms of accuracy. To address the similarity and cold start problem, a random walk algorithm combining trust and similarity are proposed, the random walk algorithm is proposed with the weight TS. The experimental results indicate that the algorithm performs better with the all user data sets and cold start data sets than others in the aspect of accuracy rate,coverage rate as well as the time complexity. The trust value of the data sets are used in the paper than an effective method to compute the value.The algorithm of this thesis improves the precision of recommendation,coverage rate and quality of recommendations.In a social network, a recommendation method based on clustering of trust is proposed, this algorithm solves the cold start problem. The experimental results indicate that the algorithms could solve the data sparseness problem and the cold start problem,the algorithms perform better than others in the aspect of accuracy rate,coverage rate.