Research map / Time series / Long-term forecasting
Are Transformers Effective for Time Series Forecasting?
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use a one-layer temporal linear model for direct multi-step forecasting | LTSF forecasting · Nine real-life datasets | MSE: 20\% $\sim$ 50\% | Helped |
| Replace Informer attention layers with linear layers (ablation) | Informer · Exchange-Rate, horizon 96 | MSE: 0.847 -> 1.003 | Hurt |
| Remove Informer auxiliary designs, keeping embeddings and a linear layer (ablation) | Informer · Exchange-Rate, horizon 96 | MSE: 1.003 -> 0.173 | Helped |
| Simplify Informer to a single linear layer (ablation) | Informer · Exchange-Rate, horizon 96 | MSE: 0.173 -> 0.084 | Helped |
| Remove positional embeddings (ablation) | FEDformer · Traffic, horizon 96 | MSE: 0.597 -> 0.587 | Helped |
| Remove timestamp embeddings (ablation) | FEDformer · Traffic, horizon 96 | MSE: 0.597 -> 0.613 | Hurt |
| Remove timestamp embeddings (ablation) | Autoformer · Traffic, horizon 96 | MSE: 0.629 -> 0.681 | Hurt |
| Remove positional embeddings (ablation) | Informer · Traffic, horizon 96 | MSE: 0.719 -> 1.035 | Hurt |
| Train FEDformer on shortened rather than full Traffic training data (ablation) | FEDformer · Traffic, horizon 96 | MSE: 0.587 -> 0.568 | Helped |
| Use a far rather than adjacent historical input window (ablation) | FEDformer · Electricity, horizon 720 | MSE: 0.251 -> 0.265 | Hurt |
| Shuffle the raw input sequence (ablation) | Linear · Exchange-Rate | MSE: 27.26\% | Hurt |
| Shuffle the raw input sequence (ablation) | FEDformer · ETTh1 | MSE: 73.28\% | Hurt |
Closest papers by meaning
- A Closer Look at Transformers for Time Series Forecasting: Understanding Why They Work and Where They Struggle (ICML 2025 · similarity 0.91)
- Unlocking the Power of Patch: Patch-Based MLP for Long-Term Time Series Forecasting (AAAI 2025 · similarity 0.89)
- Are Self-Attentions Effective for Time Series Forecasting? (NeurIPS 2024 · similarity 0.89)
- iTransformer: Inverted Transformers Are Effective for Time Series Forecasting (ICLR 2024 · similarity 0.88)
- Sequence Complementor: Complementing Transformers for Time Series Forecasting with Learnable Sequences (AAAI 2025 · similarity 0.88)
- SAMformer: Unlocking the Potential of Transformers in Time Series Forecasting with Sharpness-Aware Minimization and Channel-Wise Attention (ICML 2024 · similarity 0.87)
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.