The project provides the following major functions:
- Analyzing historic data and training machine learning models as well as finding their best hyper-parameters. It is performed in batch off-line mode.
- Signaling service which is regularly requests new data from the exchange and generates buy-sell signals by applying the previously trained models
- Trading service which does real trading by buying or selling the assets according to the generated signals
This software is running in a cloud in test mode and sends its signals to this Telegram channel:
https://t.me/intelligent_trading_signals
Everybody can subscribe to the channel to get the impression about the signals this bot generates. What you will see are messages like this one:
₿ 60,518 📉 Score: -0.26
This score is for BTC and the value -0.26 mean negative forecast (for the next 1-3 hours). If the score is around 0 then no messages are sent. Also, the latest close price is provided, that is, it is highly likely that you will see the low price 1-2% lower than this close price.
Currently, the bot is configured using the following parameters:
- Exchange: Binance
- Cryptocurrency: ₿ Bitcoin
- Frequency: 1 minute
- Score between -1 and +1. <0 means decrease, and >0 means increase
- Filter: notifications are sent only if score is greater than ±0.15
- One increase/decrease sign is added for each step of 0.5 (after the filter threshold)
- Prediction horizon 3 hours ahead. For example, if the score is +0.25 then the price is likely to increase 1-2% during next 3 hours
- History taken into account for forecasts: 12-24 hours
Source code: https://github.com/asavinov/intelligent-trading-bot