基于一跳信任模型的协同过滤推荐算法

基于社会信任网络的协同过滤推荐算法存在节点之间多下一跳带来的复杂路径选择和信任弱传递问题。针对这2个问题，给出基于项目的一跳信任模型，该模型通过用户对项目信任度的计算，定义用户的直接和间接社会信任属性，然后一步跳转计算用户之间的直接和间接信任距离，进而计算用户之间的信任度。基于此模型设计推荐算法，同时分析了信任度与传统相似度的理论关系并二维拟合。仿真实验表明，该算法提高了推荐准确度(约0.02 MAE)，降低了训练时间(约50%)。

Collaborative filtering recommendation algorithm based on one-jump trust model

A collaborative filtering recommendation algorithm based on the trust network of social brings two problems that the choice of complex paths between nodes and the weak transfering of trust.Toward to these two problems,a one-jump trust model based on items was put forward,the model calculated the trust between users and items,defined the consumer’s trust attribute vector of social and calculated the direct and indirect distance one-jump by items,and then calculated the trust between users.A collaborative filtering algorithm(OneJ-TCF) is degined based on the model,moreover analysed and reorganized the relation between trust and similarity.The experiments show that this algorithm improves the degree of accuracy(reducing about 0.02 MAE),and saves about 50% training time at the same time.