Predicts Bitcoin price trends using an LSTM-RNN with technical indicators for automated trading via the Binance API.
LSTM Crypto Price Prediction is a machine learning project that forecasts Bitcoin price trends using an LSTM-RNN model. It analyzes historical BTC data with technical indicators like MACD and RSI to generate buy/sell signals for automated trading bots. The goal is to identify upward or downward trends to inform trading decisions through the Binance API.
Developers and data scientists interested in applying machine learning to cryptocurrency trading, particularly those building automated trading systems or experimenting with financial prediction models.
It provides an open-source, customizable framework for crypto price prediction with integrated technical analysis, real-world trading simulation, and a focus on practical market application. Developers can extend the model with additional features or reinforcement learning for improved accuracy.
Predicting price trends in cryptomarkets using an lstm-RNN for the use of a trading bot
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Integrates multiple technical indicators like MACD, Stochastic RSI, and Coppock Curve, along with polynomial interpolation, to enrich input data for the LSTM model, as detailed in the README's features section.
Includes code to simulate trades using Binance API data, accounting for exchange fees and market conditions, with example results showing performance metrics like wallet percentage changes.
Applies a Savitzky-Golay filter to smooth volatile Bitcoin price data without significant signal distortion, improving training label quality, as explained in the label generation section.
Provides modifiable code for developers to extend with additional features or reinforcement learning, as suggested in the README's results and updates.
Simulations show the bot loses money (-11.26% with fees) compared to holding Bitcoin, indicating it's not effective for actual trading without significant improvements.
The model is 'too laggard in trading decisions,' causing late trades that are costly in volatile markets, as admitted in the trading test results.
Only produces binary buy/sell signals without dynamic adjustments, which the README notes is insufficient for complex market conditions and requires a more dynamic loss function.