Time series forecasting baselines

This is a repository for time series forecasting baselines. We mainly focus on deep learning based methods, such as TCN, LSTM, Transformer based methods.

We aim to provide an environment for developers who need to reimplement SOTA paper works and to apply SOTA works into their own tasks.

Requirements

yacs (Yet Another Configuration System)
PyTorch (An open source deep learning platform)

In a Nutshell
In Details
Future Work
Contributing
Acknowledgments

In a Nutshell

In a nutshell here's how to use this code, so for example assume you want to run Informer to train ETT, so you should do the following:

bash bash_scripts/run_debug.sh

The algoroithm related codes are placed in tsf_baselines. The config and scripts related codes are placed out of tsf_baselines directory.

In Details

├── tsf_baselines - algorithm related codes.
│    └── config
│           └── defaults.py  - here's the default config file.
│
│    └── data  
│           └── datasets  - here's the datasets folder that is responsible for all data handling.
│           └── transforms  - here's the data preprocess folder that is responsible for all data augmentation.
│           └── build.py  		   - here's the file to make dataloader.
│           └── collate_batch.py   - here's the file that is responsible for merges a list of samples to form a mini-batch.
│
│     └── engine
│            └── trainer.py     - this file contains the train loops.
│            └── inference.py   - this file contains the inference process.
│
│     └── layers             - this folder contains any customed layers of your project.
│
│     └── solver             - this folder contains optimizer of your project.
│            └── build.py
│            └── lr_scheduler.py
│   
│     └── algorithm            - this folder contains any algorithm of your project.
│            └── build.py
│
│     └── modeling            - this folder contains any backbone model of your project.
│            └── informer
│     └── utils
│            └── logger.py
│            └── any_other_utils_you_need
│
├── configs  
│    └── informer
│           └── train_informer_v1.yml  - here's the specific config file for specific model or dataset.
│ 
├── tools                - here's the train/test model of your project.
│    └── train_net.py   - here's an example of train model that is responsible for the whole pipeline.
│ 
└── tests					- this foler contains unit test of your project.
     ├── test_data_sampler.py

Results

All the results are the average of 5 runs in the same setting.

Informer - univariate forecasting

Dataset	Setting	MSE	MAE	RMSE	MAPE	MSPE
ETTm1	24	0.024	0.116	0.153	0.090	0.015
ETTm1	48	0.056	0.182	0.237	0.139	0.033
ETTm1	96	0.215	0.387	0.462	0.276	0.100
ETTm1	288	0.362	0.523	0.601	0.380	0.182
ETTm1	672	0.527	0.646	0.725	0.468	0.271

Future Work

Complete the Informer part and give a results table of this reimplemetation.
Add other deep learning based algorithms.

Contributing

Any kind of enhancement or contribution is welcomed.

Acknowledgments

Great thanks to Sidun Liu.

Good competition and collaboration experience with him. This repository is mainly built by us.

This project design is highly borrowed from Deep-Learning-Project-Template, Domainbed.

The algorithm code is highly borrowed from Informer2020

zhaoyang10 / time-series-forecasting-baselines Goto Github PK