CRAFT: Time Series Forecasting with Cross-Future Behavior Awareness
Authors: Yingwei Zhang, Ke Bu, Zhuoran Zhuang, Tao Xie, Yao Yu, Dong Li, Yang Guo, Detao Lv
IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct experiments on real-world dataset. Experiments on both offline large-scale dataset and online A/B test demonstrate the effectiveness of CRAFT. |
| Researcher Affiliation | Collaboration | Yingwei Zhang1,2, , Ke Bu3, , Zhuoran Zhuang3 , Tao Xie1,2 , Yao Yu3 , Dong Li3 , Yang Guo1,2 , Detao Lv3, 1Beijing Key Laboratory of Mobile Computing and Pervasive Device, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China 2University of Chinese Academy of Sciences, Beijing, China 3 Alibaba Group EMAIL,EMAIL |
| Pseudocode | No | The paper describes the methodology using detailed explanations and mathematical equations for the Koopman Predictor Module (KPM), Internal Trend Mining Module (ITM), and External Trend Guide Module (ETG), but it does not include a clearly labeled pseudocode or algorithm block. |
| Open Source Code | Yes | Our dataset and code are available at https://github.com/CRAFTinTSF/CRAFT. |
| Open Datasets | Yes | Our dataset and code are available at https://github.com/CRAFTinTSF/CRAFT. |
| Dataset Splits | No | The paper mentions 'We conduct offline experiments on real-world dataset collected in May 2023 at Fliggy' and discusses 'prediction window lengths K {7, 14, 30}' and 'look-back lengths T {30, 90, 180}', but it does not specify the train/test/validation split percentages or sample counts for the dataset used. |
| Hardware Specification | Yes | We conduct them on the cloud servers with two NVIDIA Tesla T4 GPUs with 16GB VRAM each. |
| Software Dependencies | Yes | All experiments are implemented with Python 3.8.5 and Pytorch 1.12.1 |
| Experiment Setup | Yes | We initialize the network parameters with Xavier Initialization [Glorot and Bengio, 2010]... the λ of ridge regression for solving the koopman matrix in the ITM module is 0.1, the number of child nodes m at the virtual hierarchy is set as 15 during hierarchical sampling. In addition, we train all models by setting the mini-batch size to 256 and using the Adam optimizer with a learning rate of 0.001. Except for MQRNN with the quantile loss at 0.5, all other models choose MSE as the training loss. The number of training epochs is 2 on the dataset, and the value of each experimental result is the average of 5 repeated tests. |