A Fair and Scalable Time Series Forecasting Benchmark and Toolkit.

English | 简体中文

---

🎉 Getting Started | 💡 Overall Design

📦 Dataset | 🛠️ Scaler | 🧠 Model | 📉 Metrics | 🏃‍♂️ Runner | 📜 Config | 📜 Baselines

$\text{BasicTS}^{+}$ (Basic Time Series) is a benchmark library and toolkit designed for time series forecasting. It now supports a wide range of tasks and datasets, including spatial-temporal forecasting and long-term time series forecasting. It covers various types of algorithms such as statistical models, machine learning models, and deep learning models, making it an ideal tool for developing and evaluating time series forecasting models.

If you find this project helpful, please don't forget to give it a ⭐ Star to show your support. Thank you!

On one hand, BasicTS provides a unified and standardized pipeline, offering a fair and comprehensive platform for reproducing and comparing popular models.

On the other hand, BasicTS offers a user-friendly and easily extensible interface, enabling quick design and evaluation of new models. Users can simply define their model structure and easily perform basic operations.

You can find detailed tutorials in Getting Started. Additionally, we are collecting ToDo and HowTo items. If you need more features (e.g., additional datasets or benchmark models) or tutorials, feel free to open an issue or leave a comment here.

Important

If you find this repository useful for your work, please cite it as such:

@article{shao2023exploring,
   title={Exploring Progress in Multivariate Time Series Forecasting: Comprehensive Benchmarking and Heterogeneity Analysis},
   author={Shao, Zezhi and Wang, Fei and Xu, Yongjun and Wei, Wei and Yu, Chengqing and Zhang, Zhao and Yao, Di and Jin, Guangyin and Cao, Xin and Cong, Gao and others},
   journal={arXiv preprint arXiv:2310.06119},
   year={2023}
 }

✨ Highlighted Features

Fair Performance Review

Users can compare the performance of different models on arbitrary datasets fairly and exhaustively based on a unified and comprehensive pipeline.

Developing with BasicTS

Minimum Code

Users only need to implement key codes such as model architecture and data pre/post-processing to build their own deep learning projects.

Everything Based on Config

Users can control all the details of the pipeline through a config file, such as the hyperparameter of dataloaders, optimization, and other tricks (*e.g.*, curriculum learning).

Support All Devices

BasicTS supports CPU, GPU and GPU distributed training (both single node multiple GPUs and multiple nodes) thanks to using EasyTorch as the backend. Users can use it by setting parameters without modifying any code.

Save Training Log

Support `logging` log system and `Tensorboard`, and encapsulate it as a unified interface, users can save customized training logs by calling simple interfaces.

🚀 Installation and Quick Start

For detailed instructions, please refer to the Getting Started tutorial.

📦 Supported Baselines

BasicTS implements a wealth of models, including classic models, spatial-temporal forecasting models, and long-term time series forecasting model:

Spatial-Temporal Forecasting

📊Baseline	📝Title	📄Paper	💻Code	🏛Venue	🎯Task
DCRNN	Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting	Link	Link1, Link2	ICLR'18	STF
STGCN	Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting	Link	Link	IJCAI'18	STF
GWNet	Graph WaveNet for Deep Spatial-Temporal Graph Modeling	Link	Link	IJCAI'19	STF
AGCRN	Adaptive Graph Convolutional Recurrent Network for Traffic Forecasting	Link	Link	NeurIPS'20	STF
MTGNN	Connecting the Dots: Multivariate Time Series Forecasting with Graph Neural Networks	Link	Link	SIGKDD'20	STF
StemGNN	Spectral Temporal Graph Neural Network for Multivariate Time-series Forecasting	Link	Link	NeurIPS'20	STF
GTS	Discrete Graph Structure Learning for Forecasting Multiple Time Series	Link	Link	ICLR'21	STF
STGODE	Spatial-Temporal Graph ODE Networks for Traffic Flow Forecasting	Link	Link	SIGKDD'21	STF
STNorm	Spatial and Temporal Normalization for Multi-variate Time Series Forecasting	Link	Link	SIGKDD'21	STF
D2STGNN	Decoupled Dynamic Spatial-Temporal Graph Neural Network for Traffic Forecasting	Link	Link	VLDB'22	STF
STEP	Pretraining Enhanced Spatial-temporal Graph Neural Network for Multivariate Time Series Forecasting	Link	Link	SIGKDD'22	STF
STID	Spatial-Temporal Identity: A Simple yet Effective Baseline for Multivariate Time Series Forecasting	Link	Link	CIKM'22	STF
DGCRN	Dynamic Graph Convolutional Recurrent Network for Traffic Prediction: Benchmark and Solution	Link	Link	ACM TKDD'23	STF
MegaCRN	Spatio-Temporal Meta-Graph Learning for Traffic Forecasting	Link	Link	AAAI'23	STF
STAEformer	Spatio-Temporal Adaptive Embedding Makes Vanilla Transformer SOTA for Traffic Forecasting	Link	Link	CIKM'23	STF
STWave	When Spatio-Temporal Meet Wavelets: Disentangled Traffic Forecasting via Efficient Spectral Graph Attention Networks	Link	Link	ICDE'23	STF

Long-Term Time Series Forecasting

📊Baseline	📝Title	📄Paper	💻Code	🏛Venue	🎯Task
Informer	Beyond Efficient Transformer for Long Sequence Time-Series Forecasting	Link	Link	AAAI'21	LTSF
Autoformer	Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting	Link	Link	NeurIPS'21	LTSF
HI	Historical Inertia: A Powerful Baseline for Long Sequence Time-series Forecasting	Link	None	CIKM'21	LTSF
Pyraformer	Low complexity pyramidal Attention For Long-range Time Series Modeling and Forecasting	Link	Link	ICLR'22	LTSF
FEDformer	Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting	Link	Link	ICML'22	LTSF
FiLM	Frequency improved Legendre Memory Model for LTSF	Link	Link	NeurIPS'22	LTSF
NSformer	Exploring the Stationarity in Time Series Forecasting	Link	Link	NeurIPS'22	LTSF
Triformer	Triangular, Variable-Specific Attentions for Long Sequence Multivariate Time Series Forecasting	Link	Link	IJCAI'22	LTSF
TimesNet	Temporal 2D-Variation Modeling for General Time Series Analysis	Link	Link	ICLR'23	LTSF
TiDE	Long-term Forecasting with TiDE: Time-series Dense Encoder	Link	Link	TMLR'23	LTSF
PatchTST	A Time Series is Worth 64 Words: Long-term Forecasting with Transformers	Link	Link	ICLR'23	LTSF
NHiTS	Neural Hierarchical Interpolation for Time Series Forecasting	Link	Link	AAAI'23	LTSF
DSformer	A Double Sampling Transformer for Multivariate Time Series Long-term Prediction	Link	Link	CIKM'23	LTSF
DLinear	Are Transformers Effective for Time Series Forecasting?	Link	Link	AAAI'23	LTSF
Crossformer	Transformer Utilizing Cross-Dimension Dependency for Multivariate Time Series Forecasting	Link	Link	ICLR'23	LTSF
NLinear	Are Transformers Effective for Time Series Forecasting?	Link	Link	AAAI'23	LTSF
Koopa	Learning Non-stationary Time Series Dynamics with Koopman Predictors	Link	Link	NeurIPS'24	LTSF
iTrainsformer	Inverted Transformers Are Effective for Time Series Forecasting	Link	Link	ICLR'24	LTSF
SparseTSF	Modeling LTSF with 1k Parameters	Link	Link	ICML'24	LTSF
TimeMixer	Decomposable Multiscale Mixing for Time Series Forecasting	Link	Link	ICLR'24	LTSF
UMixer	An Unet-Mixer Architecture with Stationarity Correction for Time Series Forecasting	Link	Link	AAAI'24	LTSF
ETSformer	Exponential Smoothing Transformers for Time-series Forecasting	Link	Link	arXiv	LTSF
LightTS	Fast Multivariate Time Series Forecasting with Light Sampling-oriented MLP	Link	Link	arXiv	LTSF
MTS-Mixers	Multivariate Time Series Forecasting via Factorized Temporal and Channel Mixing	Link	Link	arXiv	LTSF
SegRNN	Segment Recurrent Neural Network for Long-Term Time Series Forecasting	Link	Link	arXiv	LTSF

Others

📊Baseline	📝Title	📄Paper	💻Code	🏛Venue	🎯Task
LightGBM	LightGBM: A Highly Efficient Gradient Boosting Decision Tree	Link	Link	NeurIPS'17	Machine Learning
NBeats	Neural basis expansion analysis for interpretable time series forecasting	Link	Link1, Link2	ICLR'19	Deep Time Series Forecasting
DeepAR	Probabilistic Forecasting with Autoregressive Recurrent Networks	Link	Link1, Link2, Link3	Int. J. Forecast'20	Probabilistic Time Series Forecasting
WaveNet	WaveNet: A Generative Model for Raw Audio.	Link	Link 1, Link 2	arXiv	Audio

📦 Supported Datasets

BasicTS support a variety of datasets, including spatial-temporal forecasting, long-term time series forecasting, and large-scale datasets.

Spatial-Temporal Forecasting

🏷️Name	🌐Domain	📏Length	📊Time Series Count	🔄Graph	⏱️Freq. (m)	🎯Task
METR-LA	Traffic Speed	34272	207	True	5	STF
PEMS-BAY	Traffic Speed	52116	325	True	5	STF
PEMS03	Traffic Flow	26208	358	True	5	STF
PEMS04	Traffic Flow	16992	307	True	5	STF
PEMS07	Traffic Flow	28224	883	True	5	STF
PEMS08	Traffic Flow	17856	170	True	5	STF

Long-Term Time Series Forecasting

🏷️Name	🌐Domain	📏Length	📊Time Series Count	🔄Graph	⏱️Freq. (m)	🎯Task
BeijingAirQuality	Beijing Air Quality	36000	7	False	60	LTSF
ETTh1	Electricity Transformer Temperature	14400	7	False	60	LTSF
ETTh2	Electricity Transformer Temperature	14400	7	False	60	LTSF
ETTm1	Electricity Transformer Temperature	57600	7	False	15	LTSF
ETTm2	Electricity Transformer Temperature	57600	7	False	15	LTSF
Electricity	Electricity Consumption	26304	321	False	60	LTSF
ExchangeRate	Exchange Rate	7588	8	False	1440	LTSF
Illness	Ilness Data	966	7	False	10080	LTSF
Traffic	Road Occupancy Rates	17544	862	False	60	LTSF
Weather	Weather	52696	21	False	10	LTSF

Large Scale Dataset

🏷️Name	🌐Domain	📏Length	📊Time Series Count	🔄Graph	⏱️Freq. (m)	🎯Task
CA	Traffic Flow	35040	8600	True	15	Large Scale
GBA	Traffic Flow	35040	2352	True	15	Large Scale
GLA	Traffic Flow	35040	3834	True	15	Large Scale
SD	Traffic Flow	35040	716	True	15	Large Scale

📉 Main Results

See the paper Exploring Progress in Multivariate Time Series Forecasting: Comprehensive Benchmarking and Heterogeneity Analysis.

Contributors ✨

Thanks goes to these wonderful people (emoji key):

S22 🚧 💻 🐛	LMissher 💻 🐛	CNStark 🚇	Azusa 🐛	Yannick Wölker 🐛	hlhang9527 🐛	Chengqing Yu 💻
Reborn14 📖 💻	blisky-li 💻	TensorPulse 🐛

This project follows the all-contributors specification. Contributions of any kind welcome!

🔗 Acknowledgement

BasicTS is developed based on EasyTorch, an easy-to-use and powerful open-source neural network training framework.