Accurate train travel time (T-t) is crucial for the quality and reliability of rail transport services, particularly for China–Europe Railway Express (CRE), which occupies an important position in the global transportation network. Despite transfer learning being a useful technique to address the limited data in CRE train travel time prediction, it struggles with some insurmountable problems, such as the inability to handle seasonality and non-stationarity of data. Therefore, this paper proposes a novel fusion technique that combines transfer learning, wavelet transform, and meta-learning for predicting CRE travel time with a limited amount of sample data. Specifically, transfer learning is employed to overcome data limitations in constructing machine learning models for predicting CRE travel time. Meanwhile, a wavelet transform time series decomposition is designed to reveal hidden patterns in data and improve comprehensibility and predictability. For task decomposition, a multi-task meta-learning method is proposed that obtains the loss function gradient for each task and then updates model parameters to achieve the overall optimal structure. Lastly, a fusion technique model named WT_T.R2_MAML is developed to integrate the aforementioned functions. Through the analysis of actual operational data from the CRE trains, we have validated the successful integration of the WT_T.R2_MAML model. This achievement outlines a roadmap for the future implementation of fusion technologies.

中文翻译：

---

增强中欧铁路快车的列车行驶时间预测：一种基于迁移学习的融合技术


准确的火车旅行时间 （T-t） 对于铁路运输服务的质量和可靠性至关重要，尤其是对于在全球运输网络中占有重要地位的中欧铁路班列 （CRE） 而言。尽管迁移学习是解决 CRE 火车旅行时间预测中有限数据的有用技术，但它仍面临一些无法克服的问题，例如无法处理数据的季节性和非平稳性。因此，本文提出了一种新的融合技术，该技术结合了迁移学习、小波变换和元学习，用于用有限的样本数据预测 CRE 旅行时间。具体来说，迁移学习用于克服构建机器学习模型以预测 CRE 旅行时间的数据限制。同时，设计了小波变换时间序列分解，以揭示数据中隐藏的模式并提高可理解性和可预测性。对于任务分解，提出了一种多任务元学习方法，该方法获取每个任务的损失函数梯度，然后更新模型参数以实现整体最优结构。最后，开发了一个名为 WT_T.R2_MAML 的融合技术模型来集成上述功能。通过对 CRE 列车的实际运行数据分析，我们验证了 WT_T.R2_MAML 模型的成功集成。这一成就为未来实施聚变技术勾勒出路线图。