基于循环神经网络的实时偏好感知兴趣点推荐

Real-time preference-aware point-of-interest recommendationbased on recurrent neural network

鉴于现有兴趣点推荐方法未能充分考虑用户实时偏好的影响,提出一种利用循环神经网络学习用户实时偏好嵌入的兴趣点推荐方法。为建模用户实时偏好,通过融合地理信息和时间信息学习用户实时状态嵌入。考虑历史签到信息对用户偏好决策的影响,基于矩阵分解技术开发一种组合兴趣点类别信息和用户社会信息的用户偏好嵌入表示。联合用户实时需求向量和历史偏好向量,使用一种基于注意力机制的循环神经网络框架实现用户实时兴趣点偏好的预测。经大量实验验证了所提用户实时偏好预测模型的可行性,较其它现有流行的方法,查准率和查全率分别提高了8.7%和8.3%。

Since the existing point-of-interest(POI)recommendation methods do not fully consider the impact of the user’s real-time preference,a real-time preference-aware point-of-interest recommendation using the recurrent neural network(RTPAR)was proposed.Geographical information and temporal information were fused to represent the user’s real-time demand at diffe-rent time periods.Considering that the user historical check-in information played an important role in the user preference prediction,a matrix factorization-based preference learning framework fusing POI category information and user relationship information was proposed to learn the user historical preference embedding.Combined with the user’s real-time preference and historical preference vectors,a recurrent neural network framework based on an attention mechanism was used to predict the user’s real-time POI preference.The extensive experiments show that the proposed user’s real-time preference prediction model is feasible,and the precision and recall are improved by 8.7%and 8.3%respectively,compared with other state-of-the-art methods.