摘要
针对传统推荐算法未充分考虑行为序列间的动态时间间隔、语义不规则以及用户长短期兴趣相互纠缠的问题,提出一种时间感知的用户长短期兴趣特征分离推荐算法。利用用户个性化时间聚合间隔感知和时间位置多头注意力捕获长期兴趣特征,采用动态时间间隔感知和潜在意图注意力的Time-LSTM捕获短期兴趣特征,提出长短期兴趣特征分离获取方法,分别独立捕获两种时间尺度的用户兴趣,通过注意力机制自适应融合长短期兴趣特征,提高用户兴趣特征捕获准确率。实验结果表明,该算法在预测精度指标AUC和GAUC上较对比算法均有提升,消融实验也进一步验证了该算法的必要性。
To address the problems of dynamic time intervals between behavioral sequences,semantic irregularities and intertwined long-and short-term interests of users,a time-aware recommendation algorithm for user long-and short-term interest features separation was proposed.Long-term interest features were captured using user personalized time-aggregated interval perception and temporal location multi-headed attention.Time-LSTM with dynamic time interval perception and latent intent attention was used to capture short-term interest features.To capture user interests on two time scales separately and indepen-dently,a separate acquisition method for long-and short-term interest features was proposed.To improve the accuracy of user interest feature capture,long-and short-term interest features were adaptively fused through an attention mechanism.The ablation experiments also further demonstrate the need for the algorithm.
作者
吴迪
杨利君
马文莉
WU Di;YANG Li-jun;MA Wen-li(School of Information and Electrical Engineering,Hebei University of Engineering,Handan 056038,China)
出处
《计算机工程与设计》
北大核心
2024年第5期1443-1450,共8页
Computer Engineering and Design
基金
国家自然科学基金项目(62101174)
河北省自然科学基金项目(F2020402003、F2021402005)。
关键词
个性化时间聚合间隔
动态时间间隔
长短期记忆网络
注意力机制
长短期兴趣
特征分离
推荐
personalized time gathering interval
dynamic time interval
long and short-term memory networks
attention mecha-nism
long-and short-term interest
separation of features
recommended
