Time Series Forecasting on ETTh1
0.039MSEMF-CLR
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MF-CLRForecast Horizon (H)=24, Backbone=dilated convolutional2024.10 | 0.039 | 0.149 | — | — | — | — | — | — | — | — | — | |
| CaTTForecast Horizon (H)=24, Backbone=dilated convolutional2024.10 | 0.04 | 0.152 | — | — | — | — | — | — | — | — | — | |
| TS2VecForecast Horizon (H)=24, Backbone=dilated convolutional2024.10 | 0.04 | 0.153 | — | — | — | — | — | — | — | — | — | |
| TS2Vec + SoftCLTForecast Horizon (H)=24, Backbone=dilated convolutional2024.10 | 0.04 | 0.153 | — | — | — | — | — | — | — | — | — | |
| CoSTForecast Horizon (H)=24, Backbone=dilated convolutional2024.10 | 0.041 | 0.152 | — | — | — | — | — | — | — | — | — | |
| Mean+SAITSInput Type=univariate, Imputation Strategy=With Gain estimation2024.10 | 0.0522 | — | — | — | — | — | — | — | — | — | — | |
| Mean+BRITSInput Type=univariate, Imputation Strategy=With Gain estimation2024.10 | 0.0522 | — | — | — | — | — | — | — | — | — | — | |
| Mean+GPVAEInput Type=univariate, Imputation Strategy=With Gain estimation2024.10 | 0.0522 | — | — | — | — | — | — | — | — | — | — | |
| Mean+USGANInput Type=univariate, Imputation Strategy=With Gain estimation2024.10 | 0.0522 | — | — | — | — | — | — | — | — | — | — | |
| Mean+ImputeFormerInput Type=univariate, Imputation Strategy=With Gain estimation2024.10 | 0.0522 | — | — | — | — | — | — | — | — | — | — | |
| MeanInput Type=univariate, Imputation Strategy=Original2024.10 | 0.0523 | — | — | — | — | — | — | — | — | — | — | |
| Mean+MRNNInput Type=univariate, Imputation Strategy=With Gain estimation2024.10 | 0.0523 | — | — | — | — | — | — | — | — | — | — | |
| Mean+SPINInput Type=univariate, Imputation Strategy=With Gain estimation2024.10 | 0.0523 | — | — | — | — | — | — | — | — | — | — | |
| MF-CLRForecast Horizon (H)=48, Backbone=dilated convolutional2024.10 | 0.062 | 0.189 | — | — | — | — | — | — | — | — | — | |
| CaTTForecast Horizon (H)=48, Backbone=dilated convolutional2024.10 | 0.063 | 0.193 | — | — | — | — | — | — | — | — | — | |
| TS2VecForecast Horizon (H)=48, Backbone=dilated convolutional2024.10 | 0.063 | 0.193 | — | — | — | — | — | — | — | — | — | |
| TS2Vec + SoftCLTForecast Horizon (H)=48, Backbone=dilated convolutional2024.10 | 0.063 | 0.192 | — | — | — | — | — | — | — | — | — | |
| CoSTForecast Horizon (H)=48, Backbone=dilated convolutional2024.10 | 0.066 | 0.196 | — | — | — | — | — | — | — | — | — | |
| SATIS+INFInput Type=univariate, Imputation Strategy=With Influence Function2024.10 | 0.1026 | — | — | — | — | — | — | — | — | — | — | |
| SAITSInput Type=univariate, Imputation Strategy=Original2024.10 | 0.1027 | — | — | — | — | — | — | — | — | — | — | |
| CaTTForecast Horizon (H)=168, Backbone=dilated convolutional2024.10 | 0.122 | 0.269 | — | — | — | — | — | — | — | — | — | |
| CoSTForecast Horizon (H)=168, Backbone=dilated convolutional2024.10 | 0.126 | 0.273 | — | — | — | — | — | — | — | — | — | |
| ForecastPFNData Budget=502023.11 | 0.127 | — | — | — | — | — | — | — | — | — | — | |
| ForecastPFNData Budget=5002023.11 | 0.127 | — | — | — | — | — | — | — | — | — | — | |
| FEDformerData Budget=5002023.11 | 0.133 | — | — | — | — | — | — | — | — | — | — | |
| TransformerData Budget=5002023.11 | 0.139 | — | — | — | — | — | — | — | — | — | — | |
| MF-CLRForecast Horizon (H)=168, Backbone=dilated convolutional2024.10 | 0.142 | 0.297 | — | — | — | — | — | — | — | — | — | |
| CaTTForecast Horizon (H)=336, Backbone=dilated convolutional2024.10 | 0.143 | 0.296 | — | — | — | — | — | — | — | — | — | |
| TimesFM-2.5-200MStrategy=Task-Agnostic, N=642026.01 | 0.1435 | — | — | — | 0.1334 | 0.1354 | — | — | — | — | — | |
| TimesFM-2.5-200MStrategy=Task-Specific, N=642026.01 | 0.1435 | — | — | — | 0.1401 | 0.1411 | — | — | — | — | — | |
| AutoformerData Budget=5002023.11 | 0.144 | — | — | — | — | — | — | — | — | — | — | |
| InformerData Budget=5002023.11 | 0.144 | — | — | — | — | — | — | — | — | — | — | |
| TS2Vec + SoftCLTForecast Horizon (H)=168, Backbone=dilated convolutional2024.10 | 0.144 | 0.296 | — | — | — | — | — | — | — | — | — | |
| TS2VecForecast Horizon (H)=168, Backbone=dilated convolutional2024.10 | 0.145 | 0.296 | — | — | — | — | — | — | — | — | — | |
| Moirai-1.1-R-BaseStrategy=Task-Agnostic, N=642026.01 | 0.1458 | — | — | — | 0.1383 | 0.1407 | — | — | — | — | — | |
| ImputeFormer+INFInput Type=univariate, Imputation Strategy=With Influence Function2024.10 | 0.1461 | — | — | — | — | — | — | — | — | — | — | |
| Moirai-1.1-R-BaseStrategy=Task-Specific, N=642026.01 | 0.1478 | — | — | — | 0.1413 | 0.1417 | — | — | — | — | — | |
| USGAN+INFInput Type=univariate, Imputation Strategy=With Influence Function2024.10 | 0.1486 | — | — | — | — | — | — | — | — | — | — | |
| CoSTForecast Horizon (H)=336, Backbone=dilated convolutional2024.10 | 0.15 | 0.305 | — | — | — | — | — | — | — | — | — | |
| GPVAE+INFInput Type=univariate, Imputation Strategy=With Influence Function2024.10 | 0.1543 | — | — | — | — | — | — | — | — | — | — | |
| USGANInput Type=univariate, Imputation Strategy=Original2024.10 | 0.1549 | — | — | — | — | — | — | — | — | — | — | |
| ImputeFormerInput Type=univariate, Imputation Strategy=Original2024.10 | 0.1558 | — | — | — | — | — | — | — | — | — | — | |
| MeanData Budget=502023.11 | 0.158 | — | — | — | — | — | — | — | — | — | — | |
| MeanData Budget=5002023.11 | 0.158 | — | — | — | — | — | — | — | — | — | — | |
| GPVAEInput Type=univariate, Imputation Strategy=Original2024.10 | 0.1591 | — | — | — | — | — | — | — | — | — | — | |
| BRITS+INFInput Type=univariate, Imputation Strategy=With Influence Function2024.10 | 0.1637 | — | — | — | — | — | — | — | — | — | — | |
| TS2VecForecast Horizon (H)=336, Backbone=dilated convolutional2024.10 | 0.164 | 0.321 | — | — | — | — | — | — | — | — | — | |
| TS2Vec + SoftCLTForecast Horizon (H)=336, Backbone=dilated convolutional2024.10 | 0.164 | 0.321 | — | — | — | — | — | — | — | — | — | |
| MF-CLRForecast Horizon (H)=336, Backbone=dilated convolutional2024.10 | 0.165 | 0.323 | — | — | — | — | — | — | — | — | — | |
| BRITSInput Type=univariate, Imputation Strategy=Original2024.10 | 0.1692 | — | — | — | — | — | — | — | — | — | — | |
| TS2VecForecast Horizon (H)=720, Backbone=dilated convolutional2024.10 | 0.174 | 0.34 | — | — | — | — | — | — | — | — | — | |
| TS2Vec + SoftCLTForecast Horizon (H)=720, Backbone=dilated convolutional2024.10 | 0.174 | 0.342 | — | — | — | — | — | — | — | — | — | |
| LastData Budget=502023.11 | 0.176 | — | — | — | — | — | — | — | — | — | — | |
| LastData Budget=5002023.11 | 0.176 | — | — | — | — | — | — | — | — | — | — | |
| Meta-N-BEATSData Budget=502023.11 | 0.177 | — | — | — | — | — | — | — | — | — | — | |
| Meta-N-BEATSData Budget=5002023.11 | 0.177 | — | — | — | — | — | — | — | — | — | — | |
| Chronos-T5-BaseStrategy=Task-Agnostic, N=642026.01 | 0.1805 | — | — | — | 0.1394 | 0.1424 | — | — | — | — | — | |
| Chronos-T5-BaseStrategy=Task-Specific, N=642026.01 | 0.1843 | — | — | — | 0.1416 | 0.1445 | — | — | — | — | — | |
| CoSTForecast Horizon (H)=720, Backbone=dilated convolutional2024.10 | 0.188 | 0.354 | — | — | — | — | — | — | — | — | — | |
| CaTTForecast Horizon (H)=720, Backbone=dilated convolutional2024.10 | 0.189 | 0.354 | — | — | — | — | — | — | — | — | — | |
| MRNN+INFInput Type=univariate, Imputation Strategy=With Influence Function2024.10 | 0.1905 | — | — | — | — | — | — | — | — | — | — | |
| ArimaData Budget=5002023.11 | 0.2 | — | — | — | — | — | — | — | — | — | — | |
| SPINInput Type=univariate, Imputation Strategy=Original2024.10 | 0.2 | — | — | — | — | — | — | — | — | — | — | |
| MF-CLRForecast Horizon (H)=720, Backbone=dilated convolutional2024.10 | 0.201 | 0.371 | — | — | — | — | — | — | — | — | — | |
| SeasonalNaiveData Budget=502023.11 | 0.203 | — | — | — | — | — | — | — | — | — | — | |
| SeasonalNaiveData Budget=5002023.11 | 0.203 | — | — | — | — | — | — | — | — | — | — | |
| SPIN+INFInput Type=univariate, Imputation Strategy=With Influence Function2024.10 | 0.2106 | — | — | — | — | — | — | — | — | — | — | |
| MRNNInput Type=univariate, Imputation Strategy=Original2024.10 | 0.2184 | — | — | — | — | — | — | — | — | — | — | |
| TimeMoE-200MStrategy=Task-Agnostic, N=642026.01 | 0.2401 | — | — | — | 0.2353 | 0.2396 | — | — | — | — | — | |
| TimeMoE-200MStrategy=Task-Specific, N=642026.01 | 0.2448 | — | — | — | 0.239 | 0.2435 | — | — | — | — | — | |
| FreqLensPrediction Length (H)=962026.02 | 0.2724 | 0.4189 | — | 0.5219 | — | — | — | — | — | — | — | |
| D³VAEHorizon=82023.01 | 0.292 | — | 0.424 | — | — | — | — | — | — | — | — | |
| LTSM-BundlePrediction Length=962024.06 | 0.307 | 0.377 | — | — | — | — | — | — | — | — | — | |
| LTSM-BundlePrediction Horizon=96, Training Data Percentage=5%2024.06 | 0.307 | 0.377 | — | — | — | — | — | — | — | — | — | |
| LTSM-BundlePrediction Length=96, Few-shot data percentage=5%2024.06 | 0.307 | 0.377 | — | — | — | — | — | — | — | — | — | |
| FreqLensPrediction Length (H)=1922026.02 | 0.3269 | 0.4664 | — | 0.5718 | — | — | — | — | — | — | — | |
| LTSM-BundlePrediction Length=1922024.06 | 0.329 | 0.391 | — | — | — | — | — | — | — | — | — | |
| LTSM-BundlePrediction Horizon=192, Training Data Percentage=5%2024.06 | 0.329 | 0.391 | — | — | — | — | — | — | — | — | — | |
| D³VAEHorizon=322023.01 | 0.334 | — | 0.461 | — | — | — | — | — | — | — | — | |
| LTSM-BundlePrediction Length=Avg2024.06 | 0.338 | 0.403 | — | — | — | — | — | — | — | — | — | |
| LTSM-BundlePrediction Horizon=Avg, Training Data Percentage=5%2024.06 | 0.338 | 0.403 | — | — | — | — | — | — | — | — | — | |
| LTSM-BundlePrediction Length=Avg, Few-shot data percentage=5%2024.06 | 0.338 | 0.403 | — | — | — | — | — | — | — | — | — | |
| Full2024.06 | 0.339 | 0.389 | — | — | — | — | — | — | — | — | — | |
| ArimaData Budget=502023.11 | 0.341 | — | — | — | — | — | — | — | — | — | — | |
| LTSM-BundlePrediction Length=3362024.06 | 0.346 | 0.405 | — | — | — | — | — | — | — | — | — | |
| LTSM-BundlePrediction Horizon=336, Training Data Percentage=5%2024.06 | 0.346 | 0.405 | — | — | — | — | — | — | — | — | — | |
| D³VAEHorizon=642023.01 | 0.349 | — | 0.473 | — | — | — | — | — | — | — | — | |
| Time-MoE largePrediction Horizon=96, Evaluation Protocol=Zero-shot2026.03 | 0.35 | 0.382 | — | — | — | — | — | — | — | — | — | |
| Time-MoE LargeHorizon=96, Model Size=453M, Zero-shot=true2025.09 | 0.35 | 0.382 | — | — | — | — | — | — | — | — | — | |
| PatchFormerH=96, Zero-shot=true2026.01 | 0.352 | — | — | — | — | — | — | — | — | — | — | |
| Time-MoE basePrediction Horizon=96, Evaluation Protocol=Zero-shot2026.03 | 0.357 | 0.381 | — | — | — | — | — | — | — | — | — | |
| Time-MoE baseHorizon=96, Model Size=113M, Zero-shot=true2025.09 | 0.357 | 0.381 | — | — | — | — | — | — | — | — | — | |
| TimeSqueeze basePrediction Horizon=96, Evaluation Protocol=Zero-shot2026.03 | 0.359 | 0.385 | — | — | — | — | — | — | — | — | — | |
| TimeSqueeze largePrediction Horizon=96, Evaluation Protocol=Zero-shot2026.03 | 0.36 | 0.379 | — | — | — | — | — | — | — | — | — | |
| TIME-LLMPrediction Length=962024.06 | 0.362 | 0.392 | — | — | — | — | — | — | — | — | — | |
| RLinearPrediction Length=96, Historical Horizon=3362023.05 | 0.366 | 0.391 | — | — | — | — | — | — | — | — | — | |
| GTRLook-back length=96, Prediction length=962026.02 | 0.367 | 0.391 | — | — | — | — | — | — | — | — | — | |
| DiPE-LinearPrediction Length=962024.11 | 0.369 | 0.393 | — | — | — | — | — | — | — | — | — | |
| DiPE-LinearPrediction Length=962024.11 | 0.369 | 0.393 | — | — | — | — | — | — | — | — | — | |
| Super-LinearHorizon=96, Model Size=2.5M, Zero-shot=true2025.09 | 0.369 | 0.392 | — | — | — | — | — | — | — | — | — |