Time-aware Context and Feature Enhancement Sequence Recommendation

摘要

Recently, Recurrent Neural Network (RNN) leads to a paradigm shift in recommendation tasks, which achieves state-of-the-art performance compared with traditional methods. In context-aware recommendations, context-aware models are capable of getting context information based on time. However, they fail to capture the static long-term interests or the higher-order features of users' information. Motivated by the high efficiency of modeling both context and sequence characteristics simultaneously, we proposed Time-aware Context and Feature Enhancement Neural Network (CAPNN) to integrate the time context features and high-order features within users' behavior sequences, which captures the long and short-term static interests of users. Extensive experiments on real-world datasets demonstrate that our proposed method outperforms several state-of-the-art methods in recommendation tasks.