Cryptocurrency trading bot using Machine Learning and Rust. Inspired by CyberPunkMetalHead/cryptocurrency-machine-learning-prediction-algo-trading.

This bot has been developed in response to another project and out of curiousity to see if I could build an improved version in Rust. The efficiency of the strategy itself hasn't been thoroughly tested. Use this program at your own risk!

The trading strategy is relatively simple:

• Fetch the last X days of hourly kline (candle) data from Binance.
• Train a machine learning model on the data. I'm using LightGBM, which is a fast gradient boosting framework that uses tree based learning algorithms. It's not perfect and predictions aren't nearly as close to reality as other solutions like recurrent neural networks (RNN) like LSTM, but in my testing it can provide a good indicator for basic market movements (price up or down), which is all I need for this strategy.
• Using the trained model, predict the current candle high price. If it's lower than the current open or close (i.e current) price, wait for the next candle and start over. Otherwise, place a buy order.
• Finally, the bot waits for the price to go up until the the prediction is reached. If the prediction isn't reached by the end of the candle, it just waits until the prediction is reached eventually.

Install Rust and clone this repository:

$ git clone https://github.com/sleeyax/ml-crypto-trading-bot.git
$ cd ml-crypto-trading-bot

Then, copy the config file and edit it accordingly (should be self-explanatory):

$ cp config.example.yaml config.yaml
$ vim config.yaml # or use any other text editor of choice to edit the config file

To run the bot in development mode, execute:

$ RUST_LOG=debug cargo run

To run the bot in production mode, execute:

$ RUST_LOG=info cargo run

You can also build a release binary with cargo build -r and copy it + your config file to a VPS or raspberry pi.