This is a trading bot that uses two types of LSTM models (Long Term Short Term Memory):
- LSTM model with a custom Attention layer attached to it in order to predict the closing price of a crypto-currency.
- A multi-value assosicate LSTM model for closing, lowest and highest price.
Furthermore, these models can be fused with a sentiment model that rates the positivity or negativity of the crypto-related news. The sentiment model affects the final prediction way less than the LSTM model but that can be adjusted through a bias in the fuse equation. The use of training and using these models on a crypto currency was to be able to study the patterns quicker than with normal stocks.
To changed the type of model being used, you can switch between stockPred (LSTM with Attention) and stockPredALSTM (Assosiciate LSTM).
The bot works via the Alpaca API and thus the API-Key and API-Secret are needed.
To run the bot use:
python main.py
Current inspiration for the architecture of the bot came from the following articles:
https://cs230.stanford.edu/projects_winter_2020/reports/32066186.pdf
https://machinelearningmastery.com/the-attention-mechanism-from-scratch/
https://link.springer.com/article/10.1007/s13042-019-01041-1