基于双重注意力机制的图神经网络模型的会话推荐

Session Recommendation Based on Graph Neural Network Model with Dual Attention Mechanism

针对会话推荐场景,同一个会话中的用户行为具有内在联系。将用户会话行为的上下文信息引入会话,并对会话中的行为建模;同时导入注意力机制,构建基于双重注意力机制的会话推荐模型,从多维视角去提取用户会话数据中可能隐含的潜在用户长期喜好与短期兴趣信息。该双重注意力机制可通过给用户不同的输入行为数据赋予不同的权重,从而达到对当前推荐任务关键信息的聚合,提升推荐效果和用户体验。所构建的模型在两个公开数据集上分别进行了实验,与基准模型相比较,该模型在各评价指数上都有所提高,证明了模型的有效性。

For the session recommendation scenario,user behaviors in the same session are intrinsically related.The context information of user session behavior is introduced into the session,and the behavior in the session is modeled.At the same time,the attention mechanism is introduced to construct a session recommendation model based on the dual attention mechanism.The long-term and short-term interest information of potential users that may be implied in the user session data can be extracted from a multi-dimensional perspective.The dual attention mechanism can give different weights to different input behavior data of users,so as to achieve the aggregation of key information of the current recommendation task,and improve the recommendation effect and user experience.The constructed model is tested on two public datasets respectively.Compared with the benchmark model,the model has improved in each evaluation index,which proves the effectiveness of the model.