融合时间上下文与长短期偏好的序列推荐模型

Sequence Recommendation Model Combining Temporal Context with Long-short Term Preference

在以往基于自注意力机制的序列推荐模型中,由于忽略上下文信息和长期行为信息,常导致模型无法捕捉到多层次的序列特征.因此提出一种融合时间上下文与长短期偏好的序列推荐模型.首先利用感知时间间隔的自注意力机制来建模用户和项目短期行为交互的关系模式,然后补充用户长期偏好来增强用户特征表示,并利用门控机制将长短期偏好表示进行融合生成用户最终兴趣表示.最后对用户交互项目间的共现模式进行建模,并结合用户最终兴趣表示进行项目推荐.在3个公开数据集上进行实验的结果验证了该模型的有效性.

The previous sequence recommendation models based on self-attention mechanism can not capture multi-level sequence features due to the neglect of context information and long-term behavior information.Therefore,a sequence recommendation model combining time context with short-term preference is proposed.Firstly,the self attention mechanism of perceived time interval is used to model the relationship mode of recent interaction between users and projects.Then the user′s long-term preference is supplemented to enhance the user′s feature representation.And the gating mechanism is used to fuse the long-term and short-term preference representations to generate the user′s final interest representation.Finally,the project recommendation is carried out by combining the co-occurrence mode between user interaction projects with the final expression of user′s final interest representation.The experimental results on three public data sets demonstrate the effectiveness of the proposed model.