Research On Personalized Recommendation Of Trust Propagation Model Under Social Network

With the rapid development of information technology and the Internet, social networking has become the important way for people’s communicating, accessing to information. And as development of the Internet, more and more information around us, which making it difficult for us to find the information that we interested in the Internet. Personalized Recommendation has become the effective way to solve the problem of information overload. The system could recommend items that people interested in, help them acquiring information of interest quickly and improve the user experience. As the same time, the system can also help businessman improve customer loyalty and increase profitability. However, there are some problems and limitations exist in the personalized recommendation algorithm. Collaborative filtering algorithm is one of the most successful recommendation algorithms, but it suffering from some problems like data sparseness, cold start, malicious attacks and poor scalability, which making the recommendation inaccuracy, and new users cannot get the recommend information.In order to solve those problems mentioned before, this paper focuses on the research and improvement of personalized recommendations. To achieve accuracy recommendation, the paper discusses the case of trust propagation in social network, proposed trust propagation model and recommended method based on social networks, and then discusses how to simplify the model. According to the social network trust model and the limitations of the traditional collaborative filtering, the paper introduced the distrust into the propagation of social network. The combination of distrust and recession factor of time makes the results of recommendation more realistic. Secondly, the paper introduces the user interest model into trust calculating, computing the similarity between user interests.Finally, we test the algorithm validity by experiment. Using the evaluation indicators—forecasting accuracy, coverage, precision which usually used in recommend system, testing the validity of traditional collaborative filtering, pure trust propagation and the algorithm we proposed.