Time Series Forecasting2018n/a
M4 Forecasting Competition
100,000 time series from diverse domains (finance, demographic, macro, micro, industry, other). Competition ran in 2018. Lower sMAPE/MASE/OWA is better.
Metrics:smapi, mase, owa
Paper / WebsiteCurrent State of the Art
TiDE
13.95
smapi
Top Models Performance Comparison
Top 10 models ranked by smapi
Best Score
13.9
Top Model
TiDE
Models Compared
10
Score Range
2.1
mase
| # | Model | Score | Paper / Code | Date |
|---|---|---|---|---|
| 1 | DLinearOpen Source THUML | 2.095 | May 2022 | |
| 2 | PatchTSTOpen Source IBM Research | 1.945 | Nov 2022 | |
| 3 | TiDEOpen Source Google | 1.94 | Apr 2023 | |
| 4 | AutoformerOpen Source THUML | 1.771 | Jun 2021 | |
| 5 | iTransformerOpen Source THUML | 1.764 | Oct 2023 | |
| 6 | FEDformerOpen Source THUML | 1.701 | Jan 2022 | |
| 7 | N-HiTSOpen Source Nixtla | 1.613 | Jan 2022 | |
| 8 | LMS-AutoTSFOpen Source Independent | 1.591 | Dec 2024 | |
| 9 | TimesNetOpen Source THUML | 1.585 | Oct 2022 | |
| 10 | N-BEATSOpen Source ServiceNow | 1.559 | May 2019 | |
| 11 | TimeMixerOpen Source Fudan University | 1.559 | May 2024 | |
| 12 | ES-RNNOpen Source Uber Technologies | 1.536 | A hybrid method of exponential smoothing and recurrent neural networks for time series forecastingCode | Jan 2020 |
| 13 | TimeMixer++Open Source Fudan University | 1.487 | Oct 2024 |
owa
| # | Model | Score | Paper / Code | Date |
|---|---|---|---|---|
| 1 | DLinearOpen Source THUML | 1.051 | May 2022 | |
| 2 | TiDEOpen Source Google | 1.02 | Apr 2023 | |
| 3 | PatchTSTOpen Source IBM Research | 0.998 | Nov 2022 | |
| 4 | AutoformerOpen Source THUML | 0.939 | Jun 2021 | |
| 5 | iTransformerOpen Source THUML | 0.929 | Oct 2023 | |
| 6 | FEDformerOpen Source THUML | 0.918 | Jan 2022 | |
| 7 | N-HiTSOpen Source Nixtla | 0.861 | Jan 2022 | |
| 8 | N-BEATSOpen Source ServiceNow | 0.855 | May 2019 | |
| 9 | LMS-AutoTSFOpen Source Independent | 0.854 | Dec 2024 | |
| 10 | TimesNetOpen Source THUML | 0.851 | Oct 2022 | |
| 11 | TimeMixerOpen Source Fudan University | 0.840 | May 2024 | |
| 12 | ES-RNNOpen Source Uber Technologies | 0.821 | A hybrid method of exponential smoothing and recurrent neural networks for time series forecastingCode | Jan 2020 |
| 13 | TimeMixer++Open Source Fudan University | 0.821 | Oct 2024 |
smapiPrimary
| # | Model | Score | Paper / Code | Date |
|---|---|---|---|---|
| 1 | TiDEOpen Source Google | 13.95 | Apr 2023 | |
| 2 | DLinearOpen Source THUML | 13.639 | May 2022 | |
| 3 | PatchTSTOpen Source IBM Research | 13.152 | Nov 2022 | |
| 4 | AutoformerOpen Source THUML | 12.909 | Jun 2021 | |
| 5 | FEDformerOpen Source THUML | 12.84 | Jan 2022 | |
| 6 | iTransformerOpen Source THUML | 12.684 | Oct 2023 | |
| 7 | N-HiTSOpen Source Nixtla | 11.927 | Jan 2022 | |
| 8 | LMS-AutoTSFOpen Source Independent | 11.871 | Dec 2024 | |
| 9 | N-BEATSOpen Source ServiceNow | 11.851 | May 2019 | |
| 10 | TimesNetOpen Source THUML | 11.829 | Oct 2022 | |
| 11 | TimeMixerOpen Source Fudan University | 11.723 | May 2024 | |
| 12 | TimeMixer++Open Source Fudan University | 11.448 | Oct 2024 | |
| 13 | ES-RNNOpen Source Uber Technologies | 11.374 | A hybrid method of exponential smoothing and recurrent neural networks for time series forecastingCode | Jan 2020 |
Related Papers12
TimeMixer++: A General Time Series Pattern Machine for Universal Predictive Analysis
Oct 2024Models: TimeMixer++
TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting
May 2024Models: TimeMixer
iTransformer: Inverted Transformers Are Effective for Time Series Forecasting
Oct 2023Models: iTransformer
Long-term Forecasting with TiDE: Time-series Dense Encoder
Apr 2023Models: TiDE
A Time Series is Worth 64 Words: Long-term Forecasting with Transformers
Nov 2022Models: PatchTST
TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis
Oct 2022Models: TimesNet
Are Transformers Effective for Time Series Forecasting?
May 2022Models: DLinear
N-HiTS: Neural Hierarchical Interpolation for Time Series Forecasting
Jan 2022Models: N-HiTS
FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting
Jan 2022Models: FEDformer
Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting
Jun 2021Models: Autoformer
N-BEATS: Neural basis expansion analysis for interpretable time series forecasting
May 2019Models: N-BEATS