Transformer for time series data forecasting

No data included in the repo. However, it should work fine with most csv data input.

I am considering include the data later when the repo is ready

If you like this repo, please consider give it a star

😀

Usage

Three main things:

• preprocessing.py
• main_transformer_forecast.py
• main_transformer_classifier.py

The fnn is mainly for fun.

The rest of the code are mainly helper

Training image

Dynamically shows the training process during the training

Kernel PCA

Correlation

Training

Training start

Training end

Validation

Validation start

Validation end

Folder structure

.
├── config
│   └── config.json
├── data
├── LICENSE
├── model
├── quick_setup_gpu.bash
├── range
│   └── transformer_forecast_512
├── README.md
├── requirements-gpu.txt
├── requirements.txt
├── src
│   ├── da_rnn.py
│   ├── fnn.py
│   ├── helper.py
│   ├── main_dft.py
│   ├── main_efficiency_fore_tester.py
│   ├── main_fnn.py
│   ├── main_kernel_pca.py
│   ├── main_preprocessing.py
│   ├── main_transformer_classifier.py
│   ├── main_transformer_forecast.py
│   ├── png_to_video.sh
│   ├── rotating_donut
│   ├── rotating_donut.c
│   ├── settings.py
│   ├── transformer.py
│   └── utils.py
└── visual