Research map / Time series / Long-term forecasting
Towards Robust Real-World Multivariate Time Series Forecasting: A Unified Framework for Dependency, Asynchrony, and Missingness
Jinkwan Jang, Hyungjin Park, Jinmyeong Choi, Taesup Kim
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add cross-channel attention (ablation) | CTF · SolarWind (m=0.375) | CMSE: 0.474 -> 0.452 | Improved |
| add frequency-based dynamic patching (ablation) | CTF · SolarWind (m=0.375) | CMSE: 0.494 -> 0.452 | Improved |
| add patch masking (ablation) | CTF · SolarWind (m=0.375) | CMSE: 0.458 -> 0.452 | Improved |
| use two channel tokens instead of one (ablation) | CTF · SolarWind | CMSE: 0.404 -> 0.400 | Improved |
| use three channel tokens instead of two (ablation) | CTF · ETT1 | CMSE: 0.398 -> 0.411 | Hurt |
| use one channel token instead of three (ablation) | CTF · CHS | CMSE: 0.295 -> 0.282 | Improved |
| use three channel tokens instead of one (ablation) | CTF · EPA | CMSE: 0.775 -> 0.774 | Improved |
| raise training masking ratio from 0.1 to 0.4 (ablation) | CTF · SolarWind (m=0.375) | CMSE: 0.490 -> 0.452 | Improved |
| interpolation-free CTF versus interpolation-based TimeXer | CTF · ETT1 spectral | low-band amplitude RMSE: 0.1901 -> 0.1877 | Improved |
| shorten test input length from 576 to 288 without retraining | CTF · SolarWind | avg CMSE: 0.409 -> 0.429 | Hurt |
| replace TimeXer with CTF under 0.500 missing ratio | TimeXer · SolarWind (m=0.500) | CMSE: 0.514 -> 0.475 | Improved |
| CTF versus best baseline on Weather | TimeFilter · Weather-practical | CMSE: 0.273 -> 0.275 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- TimeXer — TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variables (NeurIPS 2024) · better on 7 of 7 shared results
- TimeFilter — TimeFilter: Patch-Specific Spatial-Temporal Graph Filtration for Time Series Forecasting (ICML 2025) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- CSformer: Combining Channel Independence and Mixing for Robust Multivariate Time Series Forecasting (AAAI 2025 · similarity 0.90)
- Rethinking Channel Dependence for Multivariate Time Series Forecasting: Learning from Leading Indicators (ICLR 2024 · similarity 0.89)
- From Similarity to Superiority: Channel Clustering for Time Series Forecasting (NeurIPS 2024 · similarity 0.89)
- TimeCHEAT: A Channel Harmony Strategy for Irregularly Sampled Multivariate Time Series Analysis (AAAI 2025 · similarity 0.88)
- CARD: Channel Aligned Robust Blend Transformer for Time Series Forecasting (ICLR 2024 · similarity 0.88)
- OLinear: A Linear Model for Time Series Forecasting in Orthogonally Transformed Domain (NeurIPS 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
