基于用户偏好的信任网络随机游走推荐模型

为了提高推荐算法评分预测的准确度，解决冷启动用户推荐问题，在TrustWalker模型基础上提出一种基于用户偏好的随机游走模型——PtTrustWalker。首先，利用矩阵分解法对社会网络中的用户、项目相似度进行计算；其次，将项目进行聚类，通过用户评分计算用户对项目类的偏好和不同项目类下的用户相似度；最后，利用权威度和用户偏好将信任细化为不同类别下用户的信任，并在游走过程中利用信任用户最高偏好类中与目标物品相似的项目评分进行评分预测。该模型降低了噪声数据的影响，从而提高了推荐结果的稳定性。实验结果表明，PtTrustWalker模型在推荐质量和推荐速度方面相比现有随机游走模型有所提高。

Trust network random walk model based on user preferences

In order to improve the accuracy of rating prediction and resolve cold-start problem in recommended systems, on the basis of the TrustWalker model, a random walk model based on user preferences, named PtTrustWalker, was proposed. Firstly, the similarities of users and items were calculated in social networks through matrix factorization method. And then, the items were clustered and the preference of user to items and the user similarity in different categories were calculated through user's scores. Finally, by making use of authority score and user preference, the credibility was detailed into user's credit in different categories, and the score was predicted by the item score of trusted user's highest preference which was similar to the target item in the process of migration. The proposed model decreases the influence of noisy data and improves the stability of the recommendation. The experimental results show that, the PtTrustWalker model has some improvements in the quality and speed of recommendation compared with the existing random walk models.