隐式反馈场景下基于Pairwise排序学习的因子分解机算法

Factorization Machines Based on Pairwise Learning with Implicit Feedback

针对隐式反馈场景下的推荐问题以及如何融入用户物品的上下文信息来进行推荐,提出了一种结合Pairwise排序学习与因子分解机的隐式反馈推荐模型。首先借鉴排序学习中Pairwise的方法解决隐式反馈负反馈缺失的问题,然后选择因子分解机作为排序函数来建模用户的上下文信息,从优化物品排序的角度出发建模用户偏好,进而针对不同用户进行个性化推荐。最终实验也表明,所提出模型在排序指标MAP和NDCG上都要优于其他3种对比算法,在解决隐式反馈下推荐问题的同时,可以利用用户的上下文信息进一步提高推荐的准确度。

In view of the problem of personalized recommendation for implicit feedback and how to incorporate users＇ contextual information in recommendation,a recommendation model combined with pairwise learning and factorization machine was proposed. First of all,the method of pairwise learning was used to solve the problem of negative feedback missing under implicit feedback scenario,and then choose the factorization machine as ranking function to model the user＇s contextual information. Specifically,the user＇s preference was modeled from the perspective of optimizing the ranking of items,eventually,provided personalized recommendations for different users according to the model score. Experiments also show that the model proposed is better than the other three contrast algorithms in terms of ranking indicators such as MAP and NDCG. While addressing the recommended problems under implicit feedback,the proposed model can further improve the accuracy of recommendation by using the user＇s contextual information.