Recurrent Translation-Based Network for Top-N Sparse Sequential Recommendation
Recurrent Translation-Based Network for Top-N Sparse Sequential Recommendation
复制标题
DOI：
10.1109/access.2019.2941083
复制
发表时间：
2019-09
期刊：
影响因子：
3.9
通讯作者：
Nuttapong Chairatanakul;T. Murata;Xin Liu
中科院分区：
文献类型：
--
作者：
Nuttapong Chairatanakul;T. Murata;Xin Liu
满足用户的需求和提高推荐系统的留存率是一个挑战。大多数用户在大多数系统中消费了一些项目。基于平移的模型在稀疏数据集上表现良好。然而，对于下一项目的用户建议，仅考虑用户和单个先前项目。或者，递归神经网络利用顺序依赖性，但在稀疏数据集上表现不佳。我们将两者统一起来，提出了基于递归翻译的网络（RTN）。RTN利用用户消费项目的序列，而不将项目之间的交互限制为最近的项目。在真实数据集上进行的实验结果表明，RTN在稀疏数据集上的性能优于其他最先进的方法。
Fulfilling users’ needs and increasing the retention rate of recommendation systems are challenging. Most users have consumed a few items in most systems. Translation-based model performs well on sparse datasets. However, a user and only single previous item are considered for the user suggestion of next items. Alternatively, recurrent neural network utilizes sequential dependency but performs poorly on sparse datasets. We unify both and propose Recurrent Translation-based Network (RTN). RTN utilizes sequences of users’ consumed items without limiting interactions between items to the most recent one. The results of conducting experiments on real-world datasets show that RTN outperforms other state-of-the-art approaches on sparse datasets.