| 摘 要: 针对深度递归方法在轨迹数据稀缺时难以捕捉用户需求多样性和轨迹不确定性的问题,基于BiGRU提出了轨迹自监督学习的城市旅行推荐方法(SLTR)。该方法通过自监督学习优化POI表示和轨迹表示,捕捉POI语义关系,并增强查询与轨迹推理能力。在4个真实城市数据集上进行对比实验,F1和Pairs-F1指标与次优模型相比,平均改进分别为2.08%和2.07%。结果表明,本文方法提高了旅行兴趣点推荐的准确度。 |
| 关键词: 旅行推荐 自监督学习 POI表示 轨迹表示 |
|
中图分类号: TP391
文献标识码: A
|
|
| Urban Travel Recommendation A Method Based on Trajectory Self-Supervised Learning |
|
ZHU Jie, GU Yiran
|
(College of Automation & College of Artificial Intelligence, Nanjing University of Posts and Telecommunications, Nanjing 210023, China)
1222056230@njupt.edu.cn; guyr@njupt.edu.cn
|
| Abstract: Addressing the challenge of deep recurrent methods in capturing user demand diversity and trajectory uncertainty under sparse trajectory data, this paper proposes a Trajectory Self-supervised Learning-based Urban Travel Recommendation method (SLTR) built upon BiGRU. The approach optimizes POI representations and trajectory representations through self-supervised learning, models POI semantic relationships, and enhances query-trajectory reasoning capabilities. Comparative experiments on four real-world urban datasets demonstrate average improvements of 2.08% and 2.07% in F1 and Pairs-F1 metrics respectively over the second-best model. Results indicate that our method enhances the accuracy of travel point-of-interest recommendations. |
| Keywords: travel recommendation sel-f supervised learning POI representation trajectory representation |
