融合信任隐性影响和信任度的推荐模型

Recommendation Model Combining Implicit Influence of Trust with Trust Degree

现有一些方法通过结合传统推荐技术和社交信息,缓解推荐系统中的冷启动问题,但由于可用的社交信息较少,效果不佳.因此,文中提出融合信任隐性影响和信任度的推荐模型,在引入社交信息中信任关系的前提下,不仅考虑用户在信任关系中的显式行为数据,还考虑信任关系的隐性影响(如被信任用户的潜在特征向量),用于获取冷启动用户的偏好特征,有效缓解在社交信息较少时不能准确地为冷启动用户进行推荐这一问题.此外,文中提出综合信任度的度量,体现目标用户与信任用户之间不同的社交影响,发挥信任用户的积极影响,提升推荐系统的性能.在3个常用数据集上的实验表明,文中方法推荐精度较高.

Some methods alleviate the cold start problem in recommender systems by combining traditional recommendation techniques and social information.However,the effect is poor due to the less available social information.Therefore,a recommendation model combining implicit influence of trust and trust degree(RIITD)is proposed in this paper.On the premise of introducing the trust relationship in social information,both the explicit behavior data of the user in the trust relationship and the implicit influence of trust relationship,such as the potential feature vector of trusted users,are taken into account to obtain the preference characteristics of cold start users.Consequently,the problem of inaccurate recommen-dation for the cold start users caused by less social information is alleviated.Moreover,the compre-hensive trust degree is introduced to reflect the different social influences between the target user and the trusted users,make the trusted users play a positive impact and improve the performance of the recommender system.Experiments on 3 commonly used datasets show that the proposed method can achieve high accuracy.