HFformer is a Transformed-based deep learning model for high-frequency trading time series forecasting. Limit Order Book (LOB) data for the USDT-BTC trading pair is collected from Binance APIs.
The MSc thesis can be accessed here.
The following Github repositories were used to build the deep learning models:
- LSTM: https://github.com/lkulowski/LSTM_encoder_decoder
- Transformer: https://github.com/AIStream-Peelout/flow-forecast
- Autoformer and FEDformer: https://github.com/cure-lab/DLinear
- TFT: https://github.com/mattsherar/Temporal_Fusion_Transform
- Spiking activation: https://github.com/nengo/pytorch-spiking
Data repository: