Reseach On The Bipartite Graph Recommendation Algorithm Based On Trust

With the development of the information age and the wide application of the Internet, E-commerce is growing up rapidly and become a part of people’s lives. More and more merchants and consumers are involved in the field of E-commerce. On the one hand, the merchants are eager to enhance their competitiveness and increase the viscosity of customers to increase profits. On the other hand, the consumers wish to buy what they really need without scanning lots of goods when they face the pages full of all kinds of pictures. Recommendation system is an ideal solution for these requirements. So recommendation algorithm becomes hot in Academic research.So far, many achievements have been made in the research of recommendation algorithm, but there is still a lot of work to do. Emerging in recent years, the bipartite graph recommendation algorithm has better performance than the collaborative filtering recommendation algorithm in personalized recommendation which is widely used in many fields. So, this thesis is designed to improve recommendation accuracy of the bipartite graph recommendation algorithm with taking use of the concept of “trust” which characterizes the user’s preference.After studying and summary about the field, this thesis is going to improve the algorithm in three aspects: building neighbor network, differentiating initial allocation resources of goods, weighting the resource transmission path.When building neighbor network, we put forward with a method that is based on trust relationship combining the user’s similarity and user’s trust which can solve the date sparsity problem. Then, we design experiments to determine the best coefficient of the combination of user’s similarity and user’s trust.When differentiating initial allocation resources of goods, we use commodity’s “degree” to measure its popularity. After the standardizing of the number of all commodities’ “degree”, the result will be used as initial allocation resource. This ideas means to get recommended degree of commodities based on their popularity.When weighting the resource transmission paths, we assign different values to describe the credibility of users’ recommendation based on the trust matrix achieved in building neighbor network. This improvement may enhance the accuracy and effectiveness of this recommendation algorithm.In the end, this article designs 3 experiments to verify the effectiveness of the algorithm by using the classic data sets. The experimental results show that “the bipartite graph recommendation algorithm based on trust”(TBG) has an obvious improvement in the accuracy of recommendation.